WO1998045775A1 - Knowledge-based information retrieval system - Google Patents

Knowledge-based information retrieval system Download PDF

Info

Publication number
WO1998045775A1
WO1998045775A1 PCT/CA1998/000306 CA9800306W WO9845775A1 WO 1998045775 A1 WO1998045775 A1 WO 1998045775A1 CA 9800306 W CA9800306 W CA 9800306W WO 9845775 A1 WO9845775 A1 WO 9845775A1
Authority
WO
WIPO (PCT)
Prior art keywords
case
attributes
attribute
new problem
candidate
Prior art date
Application number
PCT/CA1998/000306
Other languages
French (fr)
Inventor
Kalyan Moy Gupta
Alan Mark Langley
John Yen Ching
Original Assignee
1215627 Ontario Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 1215627 Ontario Inc. filed Critical 1215627 Ontario Inc.
Priority to AU69137/98A priority Critical patent/AU6913798A/en
Priority to GB9922102A priority patent/GB2338097B/en
Publication of WO1998045775A1 publication Critical patent/WO1998045775A1/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/04Inference or reasoning models

Definitions

  • the present invention relates to knowledge-based decision support system for solving problems, and more particularly, to systems using case-based reasoning, that search for and present stored solution cases that most closely relate to a new problem.
  • the traditional form of knowledge-based technology is paper documents that contain a variety of knowledge such as facts, rules, procedures, designs, and troubleshooting and problem-solving methods.
  • the contents of the paper documents cannot be manipulated, are difficult to maintain, and often only accessible to a limited number of specialists.
  • the rules and procedures are encoded and manipulated by computer programs.
  • This form of knowledge-based technology that offers decision support using rules is called a rule-based expert system.
  • CBR case-based reasoning
  • the present invention is directed to a method for assisting a user in solving a new problem case within a selected domain.
  • the method comprises the steps of providing a database comprising global domain knowledge relating to components of the selected domain, local domain knowledge, and a plurality of previously solved cases in the selected domain, each of the previously solved cases including a plurality of case attributes, said case attributes comprising case attribute names and known values associated therewith, said local domain knowledge comprising associations between the case attributes of a previously solved case; prompting the user to select a component of the domain and to select from the case attributes a set of attributes considered to be relevant to the new problem case and to provide current values for each of the new problem case attributes; searching the database of previously solved cases for candidate solved cases that include one or more of the new problem case attributes selected by the user and generating a list of said candidate solved cases; matching the candidate solved cases to the new problem case by comparing the current value for each of the new problem case attributes to the known value for the same case attribute in each of the candidate solved cases; ranking the candidate solved cases in order of
  • the local domain knowledge preferably comprises importance factors for the case attributes within a previously solved case, the importance factors being utilized in determining of which attributes questions should first be asked, precedent constraints linking case attributes within a previously solved case, the precedent constraints enabling questions related to the unanswered attributes to be generated only if the precedent constraints are satisfied, and match operators which enable values for case attributes relating to the new problem case to be matched with the known values of previously solved cases.
  • the invention is also directed to a computer system for assisting a user in solving a new problem case relating to a domain.
  • the system comprises storage means for storing local domain knowledge and previously solved case records in a database.
  • Each of said previously solved case records comprising a plurality of case attribute fields, said case attribute fields comprising case attribute names and associated values.
  • the local domain knowledge comprises associations between the case attributes of a previously solved case.
  • the system also comprises interface means for interfacing with the user, comprising output means for outputting to the user a list of case attributes of the previously solved case records, and input means for enabling the user to select from the list of case attributes a set of problem case attributes considered to be relevant to the problem case, and to input current values for case attributes relating to a new problem case, and processing means coupled to the storage means and the interface means for processing the current values of the problem case attributes.
  • the processing means comprises searching means for searching the previously solved cases for solution candidate cases; matching means for matching the solution candidate cases to the new problem case by comparing the current values of the problem case attributes with stored values for the same case attributes for each of the solution candidate cases; ranking means for ranking the solution candidate cases in order of relevance based upon the similarity and creating a list of solution candidate cases based upon said ranking; and question generation means for generating additional questions based upon unanswered attributes in the solution candidate cases for which values have not yet been provided by the user, to assist the user to enter additional current values for case attributes.
  • the present invention is further directed to a method for assisting a user in solving a new problem case within a selected domain, comprising the steps of providing a database comprising global domain knowledge relating to components of the selected domain, local domain knowledge, and a plurality of previously solved cases in the selected domain, each of the previously solved cases including a plurality of case attributes, said case attributes comprising case attribute names and known values associated therewith, said local domain knowledge comprising associations between the case attributes of a previously solved case, prompting the user to select a component of the domain and to select from the case attributes a set of attributes considered to be relevant to the new problem case and to provide current values for each of the new problem case attributes, searching the database of previously solved cases for candidate solved cases that include one or more of the new problem case attributes selected by the user and generating a list of said candidate solved cases, and matching the candidate solved cases to the new problem case by comparing the current value for each of the new problem case attributes to the known value for the same case attribute in each of the candidate solved cases.
  • Figure 1 is a schematic diagram of a system made in accordance with a preferred embodiment of the subject invention
  • Figure 2 is an conceptual illustration of the function of the subject invention.
  • Figure 3a and 3b is a flow chart showing the method implemented by the reasoning engine of the subject system to retrieve and present potential solution cases and questions;
  • Figure 4 is a flow chart showing the method used by the searcher of the subject system
  • Figures 5a and 5b are a flow chart showing the method used by the matcher of the subject system
  • Figure 6 is a flow chart showing the method used by the ranker of the subject system
  • Figures 7a and 7b are a flow chart showing the method used by the question generator of the subject system.
  • Figure 8 is a block diagram of the attribute type hierarchy of the subject system.
  • Figure 9 is an illustration of the SIM function definition about a previously solved case using the quad-tuple representation.
  • Figure 10 is a graph of the encoding of the "near_to" matching algorithm.
  • Figure 11 is a graph of the encoding of the "fuzzy_less_than” matching algorithm.
  • Figure 12 is a graph of the encoding of the "fuzzy_greater _than" matching algorithm.
  • Figure 13 is a graph of the encoding of the "range" matching algorithm.
  • Figure 14 is a graph of the encoding of the "greater_than" matching algorithm.
  • Figure 15 is a graph of the encoding of the "less_than” matching algorithm.
  • Case-based reasoning system 10 assists users in solving a new problem by retrieving case information on known problems and their solutions, within a particular domain, such as a complex apparatus, and comparing information about the new problem with the solved case information.
  • Case-based reasoning system 10 comprises a database 12 comprising case database 13 and global domain model 14, a reasoning engine 16 comprising a matcher 36 and a question generator 48, and a user interface 22.
  • Case database 13 includes for each of a plurality of previously solved cases, various case attributes 15, and attribute values 17, as well as local domain knowledge 19 in the form of attribute importance factors 21, precedence constraints 23 and match operators 25.
  • Case database 13 may be altered and new cases added without affecting the existing case history. Each case stores only the information relevant to that case, thus there is no record of domain knowledge in the case beyond the components relevant to the case. For example, if a new jet engine component is added to a domain of jet aircraft, the existing cases would not have to be updated as they do not apply to the new engine type nor is it required that they store information about it. This flexibility is achieved by utilizing a third- normal form relational database for the storage of case and domain information in separate tables.
  • the case and domain data is preferably stored in a SQL-92 compliant relational database.
  • the database engine may be a Borland Interbase Server bundled with a Delphi 2.0 Developer, although other SQL-92 compliant database servers, such as Oracle, Sybase or MicroSoft SQL Server, can be used.
  • Database 12 includes system administration tables, domain tables, case tables, and problem report tables.
  • the system adminstration tables are typically used by the case administrator to maintain meta level database control, and to control data access, grants of user rights, etc.
  • the domain tables provide the necessary descriptive language to represent a case, and typically include equipment, component and hierarchy tables.
  • the case tables preferably include a case header table and a case detail table.
  • the problem report tables are used to record information about a new problem generated during a user session, and are similar to the case tables.
  • a user will describe a new problem by specifying an attribute and a value for the attribute.
  • a problem in the area of extrusion equipment maintenance A user knows that "the die end is leaking".
  • the goal of system 10 is to assist the user in making observations that identify the cause of the leakage.
  • the various root causes i.e. stored cases
  • the relevant question to identify the root cause could be "when was the seal last changed?", “is the seal rubbing the chill roll?", "have the tightening bolts sunk in?”, and so on.
  • Question generator 48 selects and orders such questions for presentation to the user.
  • Question generator 48 receives its input from matcher 36, case database 13, and global domain model 14, and sends its output to user interface 22.
  • Matcher 36 provides the overall similarity between a new problem case and stored cases in case database 13, case database 13 provides various pieces of local domain knowledge 19, and global domain model 14 provides the format and content of the questions.
  • domain knowledge can be specified at two levels of granularity: (1.) global level; the knowledge applies to the whole case base; and
  • Precedence constraint 23 is a place holder (i.e. representation) for this kind of local domain knowledge 19. The representation method assumes conjunction when more than one precedent is specified. This assumption is reasonable since a typical case has 6-7 attributes and a few simple dependencies. However, collectively, over a family of cases, the number of constraints can be substantial and their use can accurately filter out many irrelevant and annoying questions.
  • the Importance factor 21 is used for ordering the questions and for the overall similarity (OSIM) computation.
  • Each stored previously solved case under consideration i.e., a Candidate Solved Case
  • key observations i.e., most important
  • secondary observations i.e., of lesser importance than the key observations
  • the key observations, precedent constraints 23 permitting, should be made first followed by the next set of observations and so on.
  • Match operators 25 are definitions of which attribute matching algorithm is to be used to compare the value of that attribute to the corresponding attribute in the new problem case.
  • the result of the matching algorithm is a local (SIM) definition of similarity.
  • SIM local
  • These definitions of similarity are a kind of domain knowledge used for computing the overall similarity (OSIM) of a candidate solved case with a new problem case.
  • the overall similarity (OSIM) of candidate solved cases strongly affects the question generation process.
  • Figure 2 illustrates an overview of the problem solving process.
  • the user enters the new problem description at step 170, this initial information 172 provides a description 174.
  • the description 174 is used to form criteria 176 for selection of solved problem cases from the case database 13. Cases that match the selection criteria 176 are extracted at step 178 and then ranked at step 180. The ranked cases are presented to the user and additional questions asked at step 182. If the user is satisfied with the cases presented (step 196) then the new problem case is either a known solved case 188 or a new case which will be examined by an expert forum 192 and then entered into the case database 13 at step 194. If the user is not satisfied with the previously solved cases that have been retrieved, the answers to the questions asked results in refinement at step 198 resulting in a new description at step 174 and the new selection criteria 176 are established, thus repeating the entire process.
  • Figure 3a and 3b illustrates the steps of the method 100 carried out by the reasoning engine 16 of system 10 made in accordance with the subject invention.
  • a user through the user interface 22 first selects a component (block 26) from the case database 13.
  • a domain may consist of many components.
  • the domain of a jet aircraft may consist of components for a jet engine, hydraulics system and other components.
  • the jet engine may consist of sub-components that describe particularly complex components within the engine, for example the turbine assembly.
  • the user specifies as many case attributes 50 and their values 52 as are known for the new problem to define a new problem case 30. For example, if the user is dealing with the turbine assembly component discussed above, the user may provide values for the basic "attributes" of that component. These values will be the information recorded about the new problem case. Examples of attributes in the turbine assembly component may be operating temperature or blade fractures.
  • Each case attribute 50 has information on valid values 52 stored in the case database 13.
  • searcher module 32 For each attribute 50 in the new problem case 30, searcher module 32 searches the case database 13 to find all previously solved cases that have the case attribute 50. The cases selected are added to a list of candidate solved cases 33. Once created, the list of candidate solved cases 33 is read by searcher module 32 and any duplicate cases are deleted.
  • Matcher module 36 then reads each case in the list of candidate solved cases 33, and calculates a similarity value SIM for each case attribute 50 in common with the new problem case 30. The calculation of the value of SIM is based upon the type of the case attribute 50. Once a SIM value has been computed for each of the case attributes 50 in common with the new problem case 30 and a given candidate solved case, matcher module 36 then calculates an overall similarity OSIM. OSIM is the overall similarity between the new problem case 30 and a given candidate solved case.
  • the list of candidate solved cases 33 is updated to add the OSIM value for each candidate solved case and the new list of candidate solved cases and OSIMs 39 created as input to ranker module 40.
  • Ranker module 40 reorders the list of candidate solved cases and OSIMs 39 in decreasing order of OSIM. Then based upon a selection criteria 42 such as: first five, all, or if OSIM is greater than a certain value; a list of ranked candidate solved cases 41 is created as input to the question generator module 48.
  • Question generator module 48 reads a case from the list of ranked candidate solved cases 41, the corresponding local domain knowledge 19 and global domain knowledge 14 from the database 12, and generates questions based thereupon, in a manner hereinafter described.
  • question generator 48 builds a list of unanswered attributes, i.e. case attributes which have not yet had a value provided by the user, and have had all precedence requirements met ("LA").
  • LA precedence requirements met
  • attribute identifier the rank of the attribute (i.e. the higher the OSIM ranking of the case, the higher the ranking of the attribute)
  • attribute importance category i.e. how important is it toward the confirmation of the root cause of the existing case)
  • a vote value calculated by multiplying the importance value of the attribute with its OSIM. If an attribute is found that already exists on the LA, the vote value is increased by adding to it the value of the current attribute importance multiplied by its OSIM.
  • the LA is sorted by: OSIM rank descending, and vote descending. Questions are then posed to the user at step 43 regarding the unanswered case attributes sorted highest in the sorted LA. If the user is satisfied (block 45) with a case or cases selected by the ranker 40, the session may be terminated (block 46). If the user is not satisfied with the presented cases, the user may answer the questions displayed to the user by module 43 thereby providing more information on the new problem case. The further information provided by the user adds to the definition of the new problem case 30 and the process 100 repeats.
  • Figure 4 illustrates the steps of searcher module 32.
  • the new problem case 30 provided by the user will have a number of attributes, n.
  • Step 202 sets a counter, i, that will indicate the current attribute being searched for.
  • Step 204 checks to ensure the counter i is not greater than the number of attributes n in the new problem case 30. If all attributes in the new problem case 30 have been searched for, then the process jumps to step 212. If not all attributes have been searched for the process moves to step 206, extracts all previously solved cases from the case database 13, that have a value for the attribute i. These previously solved cases are added to the list of candidate solved cases 208 and the value of i incremented at step 210.
  • Step 212 reads the list of candidates solved cases 208 and discards any duplicate candidate solved cases, thereby creating a list of unique candidate solved cases 33.
  • Figures 5a and 5b illustrate the steps of matcher module 36.
  • the matcher 36 accepts as input the list of unique candidate solved cases 33 created by the searcher 32.
  • the list of unique candidate solved cases 33 will have a number of cases, n.
  • Step 222 sets a counter, i, that will indicate the current candidate solved case being examined.
  • Step 224 checks to ensure the counter i is not greater than the number of candidate solved cases in the list of unique candidate solved cases 33.
  • step 226 If all cases in the list of unique candidate solved cases 33 have been examined then the matcher 36 has completed its function and exits via step 226. If all cases in the list of unique candidate solved cases 33 have not been examined, then the matcher 36 proceeds to step 228.
  • the current candidate solved case i.e. case #i
  • Step 228 sets a counter, j, that will indicate the current attribute being examined in the current candidate solved case.
  • Step 230 checks to ensure the counter j is not greater than the number of attributes m in the current candidate solved case. If the current attribute j is not greater than the number of attributes m in the current candidate solved case, then step 232 is invoked.
  • Step 232 checks to see if the current attribute j is in the new problem case 30 provided by the user. If the attribute j is in the new problem case 30, then step 236 adds the attribute j to a list of attributes common to the new problem case 30 and the current candidate solved case, then increments the value of the counter j at step 234. If the attribute j of the current candidate solved case is not in the new problem case 30, then step 236 is not invoked and value of the counter j is incremented at step 234. From step 234 the matcher 36 returns to step 230.
  • step 230 determines the SIM for each attribute in the list of common attributes 236. Once each SIM has been calculated, for the attributes the current candidate solved case has in common with the new problem case 30 (i.e. the list created by step 236), then step 240 calculates the OSIM for the current candidate solved case. The OSIM and the current candidate solved case are added to the list of candidate solved cases and OSIMs 39. The value of the counter i is incremented at step 244 and the matcher 36 returns to step 224.
  • Figure 6 illustrates the steps of ranker module 40.
  • the ranker 40 accepts as input the list of candidate solved cases and OSIMs 39 created by the matcher 36.
  • the first function performed by the matcher 40 is to sort the list of candidate solved cases and OSIMs 39 in descending order of OSIM. This function is performed at step 252, which creates a list of sorted candidate solved cases and OSIMs 254.
  • a system defined selection criteria is then applied at 256 to determine which cases are to be displayed to the user and these cases are output in a list of selected and ranked candidate solved cases 41.
  • Step 260 initializes an empty list of attributes LA.
  • Step 262 initializes a counter i which will indicate the number of the current candidate solved case being examined from the list of selected and ranked solved candidate cases 41.
  • Step 264 checks to see if their are any more candidate solved cases to be examined, if there are candidate solved cases left, the process moves to step. 266.
  • Step 266 uses the case information stored in the case database 13 to create a list of attributes in the current candidate solved case that are enabled and have had their precedents met, designated as LEUA.
  • Step 268 then initializes a counter j that will be used to step through the attributes in the list LEUA.
  • Step 270 if the last attribute in list LEUA has not been examined, the process moves to step 274.
  • Step 274 checks to see if the attribute j is in the list of attributes LA. If it is not, the attribute, it's rank, importance and vote are added to list LA by step 276. If it is in the list LA, then the vote value for that attribute is incremented by adding the attribute importance multiplied by the OSIM to the current vote at step 278. Both steps 276 and 278 then proceed to step 280 where the value of j is incremented.
  • Step 280 proceeds to step 270 and the next attribute in LEUA is checked. If at step 270 all the attributes in LEUA have been examined, step 270 proceeds to step 272 where the counter i is incremented.
  • Step 272 then proceeds to step 264 where the number of the current candidate solved case in the list of selected and ranked solved candidate cases 41 is examined. If there are no more cases in the list of selected and ranked solved candidate cases, then step 264 proceeds to step 282.
  • Step 282 sorts the list of attributes LA by rank descending, importance descending and vote descending and passes the sorted list LA to step 284.
  • Step 284 checks database 12 for a question to ask for attribute at the top of the sorted list LA and poses the question to the user at step 286.
  • Figure 8 lists attribute types 300 based on their properties.
  • System 10 allows for case attributes having various types of values.
  • attribute categorization only symbolic 301 has distinct subtypes. It is the subtypes that are used to categorize and evaluate attributes. Thus, attributes may be categorized into eleven distinct types as shown below.
  • the similarity between two values of an attribute is computed by a similarity computation scheme.
  • the generally applicable (i.e., global) similarity computation scheme does not consider any contextual or local information.
  • the local or contextual information resides in the cases.
  • the global scheme is used by default. If a local scheme resides in a case it will overrule the global scheme for that particular case.
  • the system should allow disabling of local schema. This would allow a knowledge engineer to determine the impact of local schema on the quality of output produced by the system.
  • Only symbolic logical attributes do not require a similarity computation scheme because they are always exact matches. Lack of a similarity computation scheme implies exact matching.
  • a symbolic attribute can be assigned symbol/labels as values. For example, a temperature may be "high”, “medium”, or "low”.
  • a numeric attribute can be assigned numbers as values, e.g.: 1.56, or 10.
  • the symbolic nominal attribute type accepts a symbolic value.
  • the attribute CITY can be assigned a value like "Hamilton”, “Toronto”, “Guelph”, or “St. Catherines”, or an attribute ENGINE LOCATION can be assigned a value like "Left -1", “Left-2", “Right-1", or “Right-2”.
  • Symbolic nominal attributes possess the following properties: a) Default Value: The default value is the usual selection that a user makes for the attribute. For example, "Toronto" as a value for the attribute CITY. It is not necessarily the normal value. Specification of a default value is optional.
  • b) Normal Value Since the present invention is a diagnostic system, it deals primarily with deviations from normal. The system is designed to ignore normal states. Specification of this property for nominal values is optional. Nominal values typically do not have normal value settings. When this property is unspecified, the attribute is not used for matching unless it is included in the stored case.
  • Similarities between any two values of a symbolic nominal attribute may be explicitly represented in a matrix.
  • the level of similarity is specified by linguistic labels such as none, very low, low, medium, high, very high, exact. These labels can be converted to numeric values based on a linear scale, or by a non-linear scale that conforms to psychological notions of distance (See for example, adverb membership modifiers such as are used in fuzzy sets).
  • the symbolic logical attribute is a special case of Symbolic-Nominal (see the attribute type taxonomy in Figure 8).
  • a logical attribute can assume only two values. For example,
  • a multi-valued attribute allows a user to assign one or more values to the attribute. This attribute type exists solely as a user convenience. For reasoning, these values are transformed into symbolic-logical-attributes with True-False or
  • the multi-valued-attribute "Fault code” can assume values FOl, F02, F03 and so on.
  • FOl and F03 the system performs an internal translation into attribute-values "Fault code F01"-present and "Fault code F03"-present.
  • Multi-value logical-attribute-references This property specifies the list of references to symbolic logical attributes, the order in which it appears in the selection option in the user interface, and the label associated with it. For example, the attribute "Fault code”has a logical-attribute reference, comprising label "FOl", its sequence number at the interface: 1, and the associated reference logical attribute ID.
  • a multi-valued attribute is never used in case representation. Instead, the component logical attributes are used. This attribute type does not possess properties for normal value or default value.
  • Symbolic-Ordinal Values assigned to this attribute type are symbolic labels that have an implicit order.
  • the temperature of a component may be "Normal”, “Warm”, “Hot”, “Very Hot”,or “Extremely hot”. Notice that these are subjective observations and are less precise than exact measurements such as 44.5 degrees.
  • the symbolic ordinal attribute type inherits its properties from the symbolic and numeric attribute types. These include the following: a) Normal Value - as for the symbolic nominal type. b) Default Value - as for the symbolic nominal type. c) Similarity computation regular quad-tuple - as for the numeric type.
  • Order number This is a real number which indicates the relative ordering of the symbolic values. For example, Normal (1), Warm (2), Hot (3), Very hot (4), and extremely Hot (5). By default, the values are set at equal intervals. However, the knowledge engineer may override the defaults to increase or decrease the similarity between adjacent symbols.
  • the similarity computation regular quad-tuple is based on the ordinal value property.
  • Numeric This attribute can be assigned a real or integral number as a value. For example, the "Temperature” is "47.5" degrees. The following properties are available: a) Default Value: This is specified when the attribute is created, and represents a subjective estimate of the most typical attribute value. This facet is required. b) Normal Value: This indicates that the attribute may not be relevant for reasoning. Unlike its symbolic counterpart, this is a range. The range consists of an upper bound and a lower bound. For example, in the attribute "Water level” normal condition refers to any value less than 4.5 m. In this case, the lower bound is - ⁇ and the upper bound is 4.5. This facet is required and must be specified when the attribute is created.
  • Min The minimum valid value that the attribute can assume. For example, the attribute "Voltage" cannot be less than 0. If unspecified, the system will not impose a lower limit on values entered by the user.
  • Max The maximum valid value that the attribute can have. For example, based on practical limits, the attribute
  • Similarity computation regular quad-tuple This is a set of four parameters defined in standard attribute units. These four parameters define the attribute similarity function at a particular value. The attribute similarity function is used to compute similarity of two values. If unspecified, the matching is exact. For details, refer to the similarity computation schemes described later.
  • Unit this is the standard dimensional unit associated with attribute values in the case base. For example, the motor current is stored in amperes. If this property is unspecified, the attribute is considered non-dimensional (i.e., None).
  • This type of attribute is computed based on two or more numeric type attributes. For example, the percentage drop in voltage is computed as (Rated-Observed) /Rated. In this example, the representation allows comparison of equipment that do not share the same voltage ratings.
  • a numeric computed attribute inherits all numeric properties. In addition, it has the following property of Computation-formula:
  • attributes can be used with the method and apparatus of the present invention, such as: a) Time. b) Date. c) Symbolic Taxonomic. These are attributes organized in a "is-a-type of" hierarchy. For example, types of house. d) User defined. This comprises any type not covered by those specified in this document that the user needs.
  • case based reasoning consists of an attribute-by-attribute comparison of the new problem description and each solved case.
  • several matching schemes will be implemented. During case creation, a knowledge engineer will be able to select the matching scheme most suitable to the problem at hand.
  • the default_fuzzy_match algorithm corresponds to the earlier implementation's handling of symbolic matching.
  • a lookup table defined in the domain model is used to retrieve the similarity of any two values of the attribute.
  • attribute name knife edge quality attribute values: new, sharp, corroded, dull similarity table: n ew sharp corroded dull n ew 1.0 0.8 0.3 0.0 sharp 1.0 0.6 0.0 corroded 1.0 0.5 dull 1.0
  • custom_fuzzy_match may be used with nominal symbolic attributes.
  • custom_fuzzy_match The custom_fuzzy_match algorithm is essentially identical to default_fuzzy_match. However, a customized lookup table stored as part of the case is used instead of the default, global lookup table. Custom_fuzzy_match is provided for special cases in which the default_fuzzy_match table is not suitable. It is not anticipated that custom_fuzzy_match will be frequently used. Because of storage and performance penalties associated with custom_fuzzy_match, default_ fuzz y_ m atch should be used whenever possible. Custom_fuzzy_match is available for nominal symbolic attributes.
  • the range algorithm may be applied to ordered, integer, and floating point attributes.
  • the less_than algorithm returns 1.0 if the attribute value falls on or below the specified threshold, and 0.0 if it is above the threshold. For example, consider a case involving precipitation. For arguments sake, assume precipitation only occurs below 15,000 feet altitude. The following representation would be used: altitude less_than 15000.0
  • the less_than algorithm may be applied to ordered, integer, and floating point attributes.
  • the fuzzy_less_than algorithm is similar to the less_than algorithm. The only distinction is a gradual rather than abrupt transition from 1.0 to 0.0 in the similarity score at the threshold value.
  • the greater_than algorithm returns 1.0 if the attribute value falls on or above the specified threshold, and 0.0 if it is below the threshold. For example, consider a case involving a hydraulic seal leakage. The problem only occurs when the pressure differential across the seal is more than 4 atmospheres. The following representation would be used: pressure differential greater_than 4.0
  • the greater_than algorithm may be applied to ordered, integer, and floating point attributes.
  • the fuzzy_greater_than algorithm is similar to the greater_than algorithm. The only distinction is a gradual rather than abrupt transition from 1.0 to 0.0 in the similarity score at the threshold value.
  • the near_to algorithm returns 1.0 if the attribute value in the problem description exactly matches the value in the stored case.
  • the match level decreases to 0.0 as the values move apart from each other. The calculation is performed according to the following equation:
  • the near_to algorithm may be applied to integer and floating point attributes.
  • the range algorithm returns 1.0 if the attribute value falls on or within the specified limits, and 0.0 if it is outside the limits. For example, consider a case involving ice buildup on an aircraft wing. For arguments sake, assume icing only occurs between 10,000 and 15,000 feet altitude. The following representation would be used: altitude range 10000.0, 15000.0
  • the range algorithm may be applied to ordered, integer, and floating point attributes.
  • the similarity-computation schemes are classified according to the applicable attribute types.
  • the following discussion deals with the computation of attribute similarity i.e.: the mapping from two values of a particular attribute to a number in the range 0.0 to 1.0.
  • the superscript "nc” on a variable refers to the variable's value in new problem case.
  • the superscript “sc” refers to the variable's value in the solved case.
  • the label “val” refers to value.
  • Figure 9 illustrates a similarity function definition using a quad- tuple representation, utilizing four parameters.
  • Computation of numeric similarities is based on four parameters (hence, the name quad-tuple or set of four). These four parameters are a, b, c, and d.
  • the parameters are defined relative to a reference value. This reference value is usually the value in the stored case.
  • the value of a, b , c, and d are specified in the units associated with the attribute.
  • Comparison schemes based on the standard deviation information for an attribute may be implemented as well. Parameters for such schemes will be defined in standard deviation units.
  • quad-tuple parameters The interpretation of the quad-tuple parameters is as follows: a. If the new value is lower than the reference value by this amount or more it is considered completely dissimilar. For example, consider a scenario where the temperature is lower by 15 degrees, resulting in significantly difference operating characteristics. b. If the new value is lower than the reference value by this amount or less it is considered essentially identical, (i.e., the decision maker is not concerned by the difference). For example, consider a scenario where the temperature is lower by 5 degrees, but the difference does not affect the outcome. c. If the new value is greater than the reference value by this amount or less it is considered essentially identical. d. If the new value is greater than the reference value by this amount or more it is considered completely dissimilar.
  • Figures 10 through 15 illustrate special case membership functions which can be derived by using different parameter settings. Each of these special cases correspond to a local similarity match algorithm as discussed earlier.
  • the case specific scheme defines /redefines the global scheme with reference to the value used in a case. Although, this lends a great deal of flexibility for reasoning and for including context specific similarity assessment, the case specific specification should be avoided.
  • the same local similarity scheme can be used by symbolic ordinal or even numeric integer types to override other numeric computation schemes.
  • the case specific similarity scheme is used to implement the custom_fuzzy_match similarity algorithm.
  • the must match scheme enforces a match with the attribute in which it is specified. That is, if the similarity with the specified attribute value is less than a preset threshold similarity value the overall similarity of the case (OSIM) is zero.
  • OSIM overall similarity of the case
  • the must match scheme is used locally in conjunction with any of the local similarity schemes. By default, the must match scheme is not enabled.
  • the must match scheme includes a local similarity threshold value.
  • the local similarity threshold value is specified by a linguistic label indicating the level of similarity.
  • the labels used must be the same as those used in the similarity matrix.
  • the local similarity scheme says that the match of motor type with value D.C. is 1 and with value A.C. is 0.
  • the must match scheme is interpreted as: the similarity of motor type must be at least exact for the case to be considered.
  • the overall similarity (OSIM) between a new problem case and a previously solved case is computed using the various matching functions of matcher module 36.
  • Four exemplary matching functions are presented hereinbelow. It is assumed that no attribute weighting (i.e. domain knowledge or contextual knowledge) is provided with a new problem case description. This implies that the attributes of a new problem case are considered as equally weighted (e.g., 1). All the functions presented here are Nearest-Neighbour like because they deviate from the true nearest-neighbour function. The deviations include consideration of only a subset of all possible attributes for matching and the use of local importance of attributes.
  • the local importance refers to the importance of an attribute in the context of a previously solved case (i.e., importance of an attribute is dependent on a previously solved case and recorded along with it).
  • Nearest- neighbour uses global weights (i.e. importance of an attribute is the same across the whole case base).
  • Prior art systems typically implement the true nearest-neighbour, and as a result do not consider local weights, nor can work with a subset of attributes.
  • modified cosine matching function (OSIM 4) is considered to be the most sophisticated one. It performs significantly better than the nearest-neighbour when local importance along with subset of all possible attributes are used for matching. It also has the ability to consider contextual variation (i.e., difference between the importance of attributes of new problem case and a previously solved case). While this provides the ability to match contexts, it can make the matching oversensitive.
  • the full contrast modified nearest-neighbour (OSIM 1) is a close second choice.
  • ⁇ w attributes are assigned according to the CSC. For those attributes not in CSC they are equally weighted using a defined scheme.
  • OSIM relevant attributes, is ⁇ ( W ,NC)2 X( w fCSC)2 capable of complete contrast and can take into account weights from NC and CSC Table 5: OSIM computing example

Abstract

A method and system for assisting a user in solving a new problem case within a selected domain, such as a complex apparatus. The method comprises the steps of providing a case database comprising domain knowledge for the selected domain and previously solved case, each previously solved case including a plurality of case attributes, said case attributes comprising case attribute names and associated values, prompting the user to select from the case attributes a set of new problem case attributes considered to be relevant to the new problem case and to provide current values for each of the new problem case attributes, searching the database of solved cases for candidate solved cases that have one or more of the new problem case attributes selected by the user and generating a list of said candidate solved cases and matching the candidate solved cases to the new problem case by comparing the value for each of the case attributes in the new problem case to the value for the same case attribute in each of the candidate solved cases.

Description

Title: KNOWLEDGE-BASED INFORMATION RETRIEVAL SYSTEM
FIELD OF THE INVENTION
The present invention relates to knowledge-based decision support system for solving problems, and more particularly, to systems using case-based reasoning, that search for and present stored solution cases that most closely relate to a new problem.
BACKGROUND OF THE INVENTION
Competitive markets have forced the organizations of today to continuously innovate and to produce increasingly sophisticated products and services. The manufacture and service of these sophisticated products is done by specialists with a high level of education, and those who have acquired skills and knowledge through substantial field experience. Reliable and high quality service is nearly impossible without these highly trained and experienced individuals. Furthermore, the complexity of these products often transcends any one individual's area of specialization. This presents an ideal opportunity for the use of knowledge-based technology to train individuals with appropriate education and to support them thereafter. Knowledge-based technology is particularly attractive because of its ability to collect, organize, and allow access to knowledge critical to an organization. In addition, it removes the risk of knowledge loss that may result from the potential loss of a specialist.
The traditional form of knowledge-based technology is paper documents that contain a variety of knowledge such as facts, rules, procedures, designs, and troubleshooting and problem-solving methods. However, the contents of the paper documents cannot be manipulated, are difficult to maintain, and often only accessible to a limited number of specialists. With the advent of computer-based technology, the rules and procedures are encoded and manipulated by computer programs. This form of knowledge-based technology that offers decision support using rules is called a rule-based expert system.
The expert system applications in early 1980s were developed in narrow and well-defined areas such as the diagnosis of bacterial infections in blood and identification of chemical structures. These early expert systems enjoyed a fair degree of success. However, the rule-based expert system applications in complex areas failed due to the following four reasons:
1. In complex, dynamic, and evolving decision environments, such as those encountered as a result of rapidly evolving complex technology, the ability to easily add rules is critical. Addition of rules affect the decisions offered by the system in unpredictable ways.
Furthermore, as the number of rules in the system grow, the system becomes extremely sluggish and its reasoning unreliable.
2. In complex areas, specialists find it difficult articulate their knowledge in the form of rules. Consequently, the expert system cannot be built.
3. The typical expert system reasoning is stymied if the specialist is unable to provide the information needed by the system. This breakdown of reasoning is termed as the brittleness of expert systems.
4. The interaction of a specialist with the expert system typically requires him/her to answer the questions in the order posed by the system. That is, the reasoning is entirely driven by the system. This is very restrictive because the specialists frequently disagree with the systems line of questioning.
To overcome these shortcomings, a number of new knowledge-based paradigms were developed. These include neural networks, fuzzy logic, and case-based reasoning (CBR) systems. Of these, CBR technology has adequately addressed the above mentioned issues and is particularly suited to complex and dynamic decision environments. Case-based reasoning systems make use of experience (i.e., previously solved cases) to solve similar new problems. The fundamental processes of CBR include the following:
1. Describe the new problem case to be solved;
2. Retrieve previously solved cases;
3. Adapt the previously solved cases to the new problem case; 4. Stop if the new problem case has been solved or;
5. Learn or acquire more knowledge about the new problem case. Many variations of these fundamental processes can be found in the scientific literature and applications (Case-Based Reasoning. Kolodner, Janet L., 1993). For example, CBR applications to problems such as diagnosis of complex machinery require incremental reasoning and problem elaboration. The nature of these problems requires that the description be revised to include new facts and evidences till the problem is resolved. The premise is that the problem definition is closely linked to the problem solution process. In contrast, many CBR applications do not require elaboration. For example, a CBR system for robot control and a CBR system for diagnosis of heart diseases. In such applications, all the needed facts or observations are available at the outset hence the problem description is complete. One can find significant variations among CBR systems in their implementations of the fundamental CBR processes. Most of the currently available CBR applications have been developed to solve relatively simple problems in narrow and well-defined areas. This limits the variety and complexity that the CBR system has to deal with. This is still beneficial because the CBR system inherently eliminates the deficiencies of rule-base systems. However, the methodologies required to provide acceptable decision support when dealing with complex problem areas need to be much more sophisticated. In other words, significant implementation differences arise depending on the intended area of application. Likewise, the methodologies discussed herein have been particularly designed to provide a CBR system with the ability to improve decision support for complex problems.
SUMMARY OF THE INVENTION
The present invention is directed to a method for assisting a user in solving a new problem case within a selected domain. The method comprises the steps of providing a database comprising global domain knowledge relating to components of the selected domain, local domain knowledge, and a plurality of previously solved cases in the selected domain, each of the previously solved cases including a plurality of case attributes, said case attributes comprising case attribute names and known values associated therewith, said local domain knowledge comprising associations between the case attributes of a previously solved case; prompting the user to select a component of the domain and to select from the case attributes a set of attributes considered to be relevant to the new problem case and to provide current values for each of the new problem case attributes; searching the database of previously solved cases for candidate solved cases that include one or more of the new problem case attributes selected by the user and generating a list of said candidate solved cases; matching the candidate solved cases to the new problem case by comparing the current value for each of the new problem case attributes to the known value for the same case attribute in each of the candidate solved cases; ranking the candidate solved cases in order of relevance based upon their similarity and presenting a list of ranked candidate solved cases in order of relevance based upon the ranking; generating additional questions based upon unanswered attributes of the candidate solved cases for which values have not yet been provided by the user, and based upon the local domain knowledge, thereby assisting the user to select and provide values for the unanswered attributes; and repeating the above steps until the user is satisfied with the list candidate solved cases.
The local domain knowledge preferably comprises importance factors for the case attributes within a previously solved case, the importance factors being utilized in determining of which attributes questions should first be asked, precedent constraints linking case attributes within a previously solved case, the precedent constraints enabling questions related to the unanswered attributes to be generated only if the precedent constraints are satisfied, and match operators which enable values for case attributes relating to the new problem case to be matched with the known values of previously solved cases.
The invention is also directed to a computer system for assisting a user in solving a new problem case relating to a domain. The system comprises storage means for storing local domain knowledge and previously solved case records in a database. Each of said previously solved case records comprising a plurality of case attribute fields, said case attribute fields comprising case attribute names and associated values. The local domain knowledge comprises associations between the case attributes of a previously solved case. The system also comprises interface means for interfacing with the user, comprising output means for outputting to the user a list of case attributes of the previously solved case records, and input means for enabling the user to select from the list of case attributes a set of problem case attributes considered to be relevant to the problem case, and to input current values for case attributes relating to a new problem case, and processing means coupled to the storage means and the interface means for processing the current values of the problem case attributes. The processing means comprises searching means for searching the previously solved cases for solution candidate cases; matching means for matching the solution candidate cases to the new problem case by comparing the current values of the problem case attributes with stored values for the same case attributes for each of the solution candidate cases; ranking means for ranking the solution candidate cases in order of relevance based upon the similarity and creating a list of solution candidate cases based upon said ranking; and question generation means for generating additional questions based upon unanswered attributes in the solution candidate cases for which values have not yet been provided by the user, to assist the user to enter additional current values for case attributes.
The present invention is further directed to a method for assisting a user in solving a new problem case within a selected domain, comprising the steps of providing a database comprising global domain knowledge relating to components of the selected domain, local domain knowledge, and a plurality of previously solved cases in the selected domain, each of the previously solved cases including a plurality of case attributes, said case attributes comprising case attribute names and known values associated therewith, said local domain knowledge comprising associations between the case attributes of a previously solved case, prompting the user to select a component of the domain and to select from the case attributes a set of attributes considered to be relevant to the new problem case and to provide current values for each of the new problem case attributes, searching the database of previously solved cases for candidate solved cases that include one or more of the new problem case attributes selected by the user and generating a list of said candidate solved cases, and matching the candidate solved cases to the new problem case by comparing the current value for each of the new problem case attributes to the known value for the same case attribute in each of the candidate solved cases.
BRIEF DESCRIPTION OF THE DRAWINGS
The present invention will now be described, by way of example only, with reference to the following drawings, in which like reference numerals refer to like parts, and in which:
Figure 1 is a schematic diagram of a system made in accordance with a preferred embodiment of the subject invention;
Figure 2 is an conceptual illustration of the function of the subject invention.
Figure 3a and 3b is a flow chart showing the method implemented by the reasoning engine of the subject system to retrieve and present potential solution cases and questions;
Figure 4 is a flow chart showing the method used by the searcher of the subject system;
Figures 5a and 5b are a flow chart showing the method used by the matcher of the subject system; Figure 6 is a flow chart showing the method used by the ranker of the subject system; and
Figures 7a and 7b are a flow chart showing the method used by the question generator of the subject system.
Figure 8 is a block diagram of the attribute type hierarchy of the subject system.
Figure 9 is an illustration of the SIM function definition about a previously solved case using the quad-tuple representation.
Figure 10 is a graph of the encoding of the "near_to" matching algorithm.
Figure 11 is a graph of the encoding of the "fuzzy_less_than" matching algorithm.
Figure 12 is a graph of the encoding of the "fuzzy_greater _than" matching algorithm.
Figure 13 is a graph of the encoding of the "range" matching algorithm.
Figure 14 is a graph of the encoding of the "greater_than" matching algorithm.
Figure 15 is a graph of the encoding of the "less_than" matching algorithm. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
Referring to Figure 1, illustrated therein is a case-based reasoning system shown generally as 10, made in accordance with a preferred embodiment of the subject invention. System 10 assists users in solving a new problem by retrieving case information on known problems and their solutions, within a particular domain, such as a complex apparatus, and comparing information about the new problem with the solved case information. Case-based reasoning system 10 comprises a database 12 comprising case database 13 and global domain model 14, a reasoning engine 16 comprising a matcher 36 and a question generator 48, and a user interface 22. Case database 13 includes for each of a plurality of previously solved cases, various case attributes 15, and attribute values 17, as well as local domain knowledge 19 in the form of attribute importance factors 21, precedence constraints 23 and match operators 25.
Case database 13 may be altered and new cases added without affecting the existing case history. Each case stores only the information relevant to that case, thus there is no record of domain knowledge in the case beyond the components relevant to the case. For example, if a new jet engine component is added to a domain of jet aircraft, the existing cases would not have to be updated as they do not apply to the new engine type nor is it required that they store information about it. This flexibility is achieved by utilizing a third- normal form relational database for the storage of case and domain information in separate tables. The case and domain data is preferably stored in a SQL-92 compliant relational database. The database engine may be a Borland Interbase Server bundled with a Delphi 2.0 Developer, although other SQL-92 compliant database servers, such as Oracle, Sybase or MicroSoft SQL Server, can be used. Database 12 includes system administration tables, domain tables, case tables, and problem report tables. The system adminstration tables are typically used by the case administrator to maintain meta level database control, and to control data access, grants of user rights, etc. The domain tables provide the necessary descriptive language to represent a case, and typically include equipment, component and hierarchy tables. The case tables preferably include a case header table and a case detail table. The problem report tables are used to record information about a new problem generated during a user session, and are similar to the case tables.
Typically, a user will describe a new problem by specifying an attribute and a value for the attribute. Consider for example a problem in the area of extrusion equipment maintenance. A user knows that "the die end is leaking". The goal of system 10 is to assist the user in making observations that identify the cause of the leakage. The various root causes (i.e. stored cases) could be that "seal size was improper", "seal was worn", "seal had come loose" and so on. The relevant question to identify the root cause could be "when was the seal last changed?", "is the seal rubbing the chill roll?", "have the tightening bolts sunk in?", and so on. Question generator 48 selects and orders such questions for presentation to the user. Question generator 48 receives its input from matcher 36, case database 13, and global domain model 14, and sends its output to user interface 22. Matcher 36 provides the overall similarity between a new problem case and stored cases in case database 13, case database 13 provides various pieces of local domain knowledge 19, and global domain model 14 provides the format and content of the questions.
Typically, domain knowledge can be specified at two levels of granularity: (1.) global level; the knowledge applies to the whole case base; and
(2.) local level; the knowledge applies only within the context of the case to which it is attached.
Attaching the domain knowledge at a local level allows a fine degree of control on the question generation process.
Often, a question should not be asked until other questions have been answered. Such situations are characterized by constraints, dependencies or other associations between two or more attributes. These constraints can be logical or practical. For example, it is not logical to ask about repeated motor tripping unless it is known that the new problem refers to tripping. Likewise, it may not be practical to observe gear teeth damage without dismantling the gear box. Precedence constraint 23 is a place holder (i.e. representation) for this kind of local domain knowledge 19. The representation method assumes conjunction when more than one precedent is specified. This assumption is reasonable since a typical case has 6-7 attributes and a few simple dependencies. However, collectively, over a family of cases, the number of constraints can be substantial and their use can accurately filter out many irrelevant and annoying questions.
The following example illustrates the use of precedence constraints 23 in a single case (see Table 1). If A and B need to be asked before C can be asked then C has precedents A and B. The question associated with attribute C is enabled locally only if the answer to A and the answer to B are "similar enough" (e.g., match > 0.5) to their respective values in the case. To give the knowledge engineer a greater degree of control a local precedence similarity threshold is provided. Table 1: Attribute dependency example
Figure imgf000014_0001
The importance of an attribute toward the confirmation of the root-cause (hypothesis) is another form of local domain knowledge 19. The Importance factor 21 is used for ordering the questions and for the overall similarity (OSIM) computation. Each stored previously solved case under consideration (i.e., a Candidate Solved Case) has a set of key observations (i.e., most important) to confirm its root cause, and some secondary observations (i.e., of lesser importance than the key observations) that provide additional confirmatory evidence or sometimes disconfirmatory evidence toward an alternative hypothesis. For example, assume that the key observations shown in Table 1 are C and D and secondary observations are E and F. The key observations, precedent constraints 23 permitting, should be made first followed by the next set of observations and so on.
Match operators 25 are definitions of which attribute matching algorithm is to be used to compare the value of that attribute to the corresponding attribute in the new problem case. The result of the matching algorithm is a local (SIM) definition of similarity. These definitions of similarity are a kind of domain knowledge used for computing the overall similarity (OSIM) of a candidate solved case with a new problem case. The overall similarity (OSIM) of candidate solved cases strongly affects the question generation process.
Figure 2 illustrates an overview of the problem solving process. The user enters the new problem description at step 170, this initial information 172 provides a description 174. The description 174 is used to form criteria 176 for selection of solved problem cases from the case database 13. Cases that match the selection criteria 176 are extracted at step 178 and then ranked at step 180. The ranked cases are presented to the user and additional questions asked at step 182. If the user is satisfied with the cases presented (step 196) then the new problem case is either a known solved case 188 or a new case which will be examined by an expert forum 192 and then entered into the case database 13 at step 194. If the user is not satisfied with the previously solved cases that have been retrieved, the answers to the questions asked results in refinement at step 198 resulting in a new description at step 174 and the new selection criteria 176 are established, thus repeating the entire process.
Figure 3a and 3b illustrates the steps of the method 100 carried out by the reasoning engine 16 of system 10 made in accordance with the subject invention. To solve a problem using system 10, a user through the user interface 22 first selects a component (block 26) from the case database 13. A domain may consist of many components. For example the domain of a jet aircraft may consist of components for a jet engine, hydraulics system and other components. In turn, the jet engine may consist of sub-components that describe particularly complex components within the engine, for example the turbine assembly.
Once a component has been selected, the user specifies as many case attributes 50 and their values 52 as are known for the new problem to define a new problem case 30. For example, if the user is dealing with the turbine assembly component discussed above, the user may provide values for the basic "attributes" of that component. These values will be the information recorded about the new problem case. Examples of attributes in the turbine assembly component may be operating temperature or blade fractures.
During the definition of the new problem case 30, the user will be informed of valid values for a selected case attribute 50. Each case attribute 50 has information on valid values 52 stored in the case database 13.
For each attribute 50 in the new problem case 30, searcher module 32 searches the case database 13 to find all previously solved cases that have the case attribute 50. The cases selected are added to a list of candidate solved cases 33. Once created, the list of candidate solved cases 33 is read by searcher module 32 and any duplicate cases are deleted.
Matcher module 36 then reads each case in the list of candidate solved cases 33, and calculates a similarity value SIM for each case attribute 50 in common with the new problem case 30. The calculation of the value of SIM is based upon the type of the case attribute 50. Once a SIM value has been computed for each of the case attributes 50 in common with the new problem case 30 and a given candidate solved case, matcher module 36 then calculates an overall similarity OSIM. OSIM is the overall similarity between the new problem case 30 and a given candidate solved case. The list of candidate solved cases 33 is updated to add the OSIM value for each candidate solved case and the new list of candidate solved cases and OSIMs 39 created as input to ranker module 40.
Ranker module 40 reorders the list of candidate solved cases and OSIMs 39 in decreasing order of OSIM. Then based upon a selection criteria 42 such as: first five, all, or if OSIM is greater than a certain value; a list of ranked candidate solved cases 41 is created as input to the question generator module 48. Question generator module 48 reads a case from the list of ranked candidate solved cases 41, the corresponding local domain knowledge 19 and global domain knowledge 14 from the database 12, and generates questions based thereupon, in a manner hereinafter described.
As question generator module 48 reads each case from the list of ranked candidate solved cases 41, question generator 48 builds a list of unanswered attributes, i.e. case attributes which have not yet had a value provided by the user, and have had all precedence requirements met ("LA"). As each new unanswered attribute is added to the LA, the following information is stored: attribute identifier, the rank of the attribute (i.e. the higher the OSIM ranking of the case, the higher the ranking of the attribute), the attribute importance category (i.e. how important is it toward the confirmation of the root cause of the existing case) and a vote value calculated by multiplying the importance value of the attribute with its OSIM. If an attribute is found that already exists on the LA, the vote value is increased by adding to it the value of the current attribute importance multiplied by its OSIM.
Once the LA has been created, it is sorted by: OSIM rank descending, and vote descending. Questions are then posed to the user at step 43 regarding the unanswered case attributes sorted highest in the sorted LA. If the user is satisfied (block 45) with a case or cases selected by the ranker 40, the session may be terminated (block 46). If the user is not satisfied with the presented cases, the user may answer the questions displayed to the user by module 43 thereby providing more information on the new problem case. The further information provided by the user adds to the definition of the new problem case 30 and the process 100 repeats.
Figure 4 illustrates the steps of searcher module 32. The new problem case 30 provided by the user will have a number of attributes, n. Step 202 sets a counter, i, that will indicate the current attribute being searched for. Step 204 checks to ensure the counter i is not greater than the number of attributes n in the new problem case 30. If all attributes in the new problem case 30 have been searched for, then the process jumps to step 212. If not all attributes have been searched for the process moves to step 206, extracts all previously solved cases from the case database 13, that have a value for the attribute i. These previously solved cases are added to the list of candidate solved cases 208 and the value of i incremented at step 210. Control then returns to step 204 and the process iterates until all attributes in the new problem case 30 have been examined and control passes to step 212. Step 212 reads the list of candidates solved cases 208 and discards any duplicate candidate solved cases, thereby creating a list of unique candidate solved cases 33. Figures 5a and 5b illustrate the steps of matcher module 36. The matcher 36 accepts as input the list of unique candidate solved cases 33 created by the searcher 32. The list of unique candidate solved cases 33 will have a number of cases, n. Step 222 sets a counter, i, that will indicate the current candidate solved case being examined. Step 224 checks to ensure the counter i is not greater than the number of candidate solved cases in the list of unique candidate solved cases 33. If all cases in the list of unique candidate solved cases 33 have been examined then the matcher 36 has completed its function and exits via step 226. If all cases in the list of unique candidate solved cases 33 have not been examined, then the matcher 36 proceeds to step 228. The current candidate solved case (i.e. case #i) from the list of unique candidate solved cases 33 will have a number of attributes say, m. Step 228 sets a counter, j, that will indicate the current attribute being examined in the current candidate solved case. Step 230 checks to ensure the counter j is not greater than the number of attributes m in the current candidate solved case. If the current attribute j is not greater than the number of attributes m in the current candidate solved case, then step 232 is invoked. Step 232 checks to see if the current attribute j is in the new problem case 30 provided by the user. If the attribute j is in the new problem case 30, then step 236 adds the attribute j to a list of attributes common to the new problem case 30 and the current candidate solved case, then increments the value of the counter j at step 234. If the attribute j of the current candidate solved case is not in the new problem case 30, then step 236 is not invoked and value of the counter j is incremented at step 234. From step 234 the matcher 36 returns to step 230. The loop of steps 230, 232, 234 and 236 repeats until at step 230 the value of the current attribute j is greater than the number of attributes m in the current candidate solved case, then the matcher 36 moves to step 238. Step 238 determines the SIM for each attribute in the list of common attributes 236. Once each SIM has been calculated, for the attributes the current candidate solved case has in common with the new problem case 30 (i.e. the list created by step 236), then step 240 calculates the OSIM for the current candidate solved case. The OSIM and the current candidate solved case are added to the list of candidate solved cases and OSIMs 39. The value of the counter i is incremented at step 244 and the matcher 36 returns to step 224.
Figure 6 illustrates the steps of ranker module 40. The ranker 40 accepts as input the list of candidate solved cases and OSIMs 39 created by the matcher 36. The first function performed by the matcher 40 is to sort the list of candidate solved cases and OSIMs 39 in descending order of OSIM. This function is performed at step 252, which creates a list of sorted candidate solved cases and OSIMs 254. A system defined selection criteria is then applied at 256 to determine which cases are to be displayed to the user and these cases are output in a list of selected and ranked candidate solved cases 41.
Figures 7a and 7b illustrate the steps of question generator module 48. Step 260 initializes an empty list of attributes LA. Step 262 initializes a counter i which will indicate the number of the current candidate solved case being examined from the list of selected and ranked solved candidate cases 41. Step 264 checks to see if their are any more candidate solved cases to be examined, if there are candidate solved cases left, the process moves to step. 266. Step 266 uses the case information stored in the case database 13 to create a list of attributes in the current candidate solved case that are enabled and have had their precedents met, designated as LEUA. Step 268 then initializes a counter j that will be used to step through the attributes in the list LEUA. At step 270 if the last attribute in list LEUA has not been examined, the process moves to step 274. Step 274 checks to see if the attribute j is in the list of attributes LA. If it is not, the attribute, it's rank, importance and vote are added to list LA by step 276. If it is in the list LA, then the vote value for that attribute is incremented by adding the attribute importance multiplied by the OSIM to the current vote at step 278. Both steps 276 and 278 then proceed to step 280 where the value of j is incremented. Step 280 proceeds to step 270 and the next attribute in LEUA is checked. If at step 270 all the attributes in LEUA have been examined, step 270 proceeds to step 272 where the counter i is incremented. Step 272 then proceeds to step 264 where the number of the current candidate solved case in the list of selected and ranked solved candidate cases 41 is examined. If there are no more cases in the list of selected and ranked solved candidate cases, then step 264 proceeds to step 282. Step 282 sorts the list of attributes LA by rank descending, importance descending and vote descending and passes the sorted list LA to step 284. Step 284 checks database 12 for a question to ask for attribute at the top of the sorted list LA and poses the question to the user at step 286.
Figure 8 lists attribute types 300 based on their properties. System 10 allows for case attributes having various types of values. In the attribute categorization only symbolic 301 has distinct subtypes. It is the subtypes that are used to categorize and evaluate attributes. Thus, attributes may be categorized into eleven distinct types as shown below.
1) Symbolic-Nominal 305 (S)
2) Symbolic-Logical 308 (L) 3) Symbolic-Multi-valued 312 (M)
4) Symbolic-Ordinal 310 (O)
5) Numeric 302 (N)
6) Computed 311 (C) Each property controls an aspect of the attribute's behaviour during run-time. Table 2 identifies the properties applicable to each attribute type.
Table 2: Properties Applicable to Each Attribute Type
Figure imgf000022_0001
In addition to the type-specific properties described below, one property is applicable across all attribute types. This is the Global-Similarity-Computation-Scheme. The similarity between two values of an attribute is computed by a similarity computation scheme. Various types of similarity computation schemes will be presented. The generally applicable (i.e., global) similarity computation scheme does not consider any contextual or local information. The local or contextual information resides in the cases. The global scheme is used by default. If a local scheme resides in a case it will overrule the global scheme for that particular case. The system should allow disabling of local schema. This would allow a knowledge engineer to determine the impact of local schema on the quality of output produced by the system. Only symbolic logical attributes do not require a similarity computation scheme because they are always exact matches. Lack of a similarity computation scheme implies exact matching.
The two broad categories of attribute types are symbolic and numeric. A symbolic attribute can be assigned symbol/labels as values. For example, a temperature may be "high", "medium", or "low". A numeric attribute can be assigned numbers as values, e.g.: 1.56, or 10.
A discussion of each attribute type follows.
1) Symbolic-Nominal
The symbolic nominal attribute type accepts a symbolic value. For example, the attribute CITY can be assigned a value like "Hamilton", "Toronto", "Guelph", or "St. Catherines", or an attribute ENGINE LOCATION can be assigned a value like "Left -1", "Left-2", "Right-1", or "Right-2". Symbolic nominal attributes possess the following properties: a) Default Value: The default value is the usual selection that a user makes for the attribute. For example, "Toronto" as a value for the attribute CITY. It is not necessarily the normal value. Specification of a default value is optional. b) Normal Value: Since the present invention is a diagnostic system, it deals primarily with deviations from normal. The system is designed to ignore normal states. Specification of this property for nominal values is optional. Nominal values typically do not have normal value settings. When this property is unspecified, the attribute is not used for matching unless it is included in the stored case.
Similarities between any two values of a symbolic nominal attribute may be explicitly represented in a matrix. The level of similarity is specified by linguistic labels such as none, very low, low, medium, high, very high, exact. These labels can be converted to numeric values based on a linear scale, or by a non-linear scale that conforms to psychological notions of distance (See for example, adverb membership modifiers such as are used in fuzzy sets). Linear scale (approx.): None (0), very low (0.16), low
(0.33), medium (0.50), high (0.67), very high (0.83), same (1.0).
Non-linear (Sigmoid scale): For example, None (0.0),
Very low (0.1) Low (0.25 ), Medium (0.5), high(0.75), Very high
(0.9), Same (1.0). The sigmoid represents the notion that human mind tends to distinguish less at the extremes and more in the neighbourhood of average values.
2) Symbolic Logical
The symbolic logical attribute is a special case of Symbolic-Nominal (see the attribute type taxonomy in Figure 8). A logical attribute can assume only two values. For example,
True-False, On-Off, Open-Closed, In-out, Above-Below, and Present-Absent. The similarity between the two values is always zero. In other words, the matching is always exact. The symbolic logical type inherits all the properties of the symbolic nominal (i.e., default value and normal value). 3) Symbolic-multi-valued
A multi-valued attribute allows a user to assign one or more values to the attribute. This attribute type exists solely as a user convenience. For reasoning, these values are transformed into symbolic-logical-attributes with True-False or
Present-Absent values. For example, the multi-valued-attribute "Fault code" can assume values FOl, F02, F03 and so on. When the user selects values FOl and F03 the system performs an internal translation into attribute-values "Fault code F01"-present and "Fault code F03"-present.
Properties for multi-valued attribute include: a) Multi-value logical-attribute-references: This property specifies the list of references to symbolic logical attributes, the order in which it appears in the selection option in the user interface, and the label associated with it. For example, the attribute "Fault code"has a logical-attribute reference, comprising label "FOl", its sequence number at the interface: 1, and the associated reference logical attribute ID. A multi-valued attribute is never used in case representation. Instead, the component logical attributes are used. This attribute type does not possess properties for normal value or default value.
4) Symbolic-Ordinal Values assigned to this attribute type are symbolic labels that have an implicit order. For example, the temperature of a component may be "Normal", "Warm", "Hot", "Very Hot",or "Extremely hot". Notice that these are subjective observations and are less precise than exact measurements such as 44.5 degrees. The symbolic ordinal attribute type inherits its properties from the symbolic and numeric attribute types. These include the following: a) Normal Value - as for the symbolic nominal type. b) Default Value - as for the symbolic nominal type. c) Similarity computation regular quad-tuple - as for the numeric type.
One additional property is required: a) Ordinal value (Order number): This is a real number which indicates the relative ordering of the symbolic values. For example, Normal (1), Warm (2), Hot (3), Very hot (4), and extremely Hot (5). By default, the values are set at equal intervals. However, the knowledge engineer may override the defaults to increase or decrease the similarity between adjacent symbols.
During reasoning, the system uses the ordinal value. The similarity computation regular quad-tuple is based on the ordinal value property.
5) Numeric This attribute can be assigned a real or integral number as a value. For example, the "Temperature" is "47.5" degrees. The following properties are available: a) Default Value: This is specified when the attribute is created, and represents a subjective estimate of the most typical attribute value. This facet is required. b) Normal Value: This indicates that the attribute may not be relevant for reasoning. Unlike its symbolic counterpart, this is a range. The range consists of an upper bound and a lower bound. For example, in the attribute "Water level" normal condition refers to any value less than 4.5 m. In this case, the lower bound is -∞ and the upper bound is 4.5. This facet is required and must be specified when the attribute is created. c) Min: The minimum valid value that the attribute can assume. For example, the attribute "Voltage" cannot be less than 0. If unspecified, the system will not impose a lower limit on values entered by the user. d) Max: The maximum valid value that the attribute can have. For example, based on practical limits, the attribute
"Voltage" cannot be more than 10,000. If unspecified, the system will not impose an upper limit on values entered by the user. e) Similarity computation regular quad-tuple: This is a set of four parameters defined in standard attribute units. These four parameters define the attribute similarity function at a particular value. The attribute similarity function is used to compute similarity of two values. If unspecified, the matching is exact. For details, refer to the similarity computation schemes described later. f) Unit: this is the standard dimensional unit associated with attribute values in the case base. For example, the motor current is stored in amperes. If this property is unspecified, the attribute is considered non-dimensional (i.e., None).
6) Computed
This type of attribute is computed based on two or more numeric type attributes. For example, the percentage drop in voltage is computed as (Rated-Observed) /Rated. In this example, the representation allows comparison of equipment that do not share the same voltage ratings. A numeric computed attribute inherits all numeric properties. In addition, it has the following property of Computation-formula:
This facet contains the function form that uses the numeric attribute references as its parameters. For example, computed value for percentage drop in voltage = (Rated-Observed)xl00/Rated.
All attributes referenced in a computed attribute must have been defined before the computed attribute is created.
Other types of attributes can be used with the method and apparatus of the present invention, such as: a) Time. b) Date. c) Symbolic Taxonomic. These are attributes organized in a "is-a-type of" hierarchy. For example, types of house. d) User defined. This comprises any type not covered by those specified in this document that the user needs.
At its most basic level, case based reasoning consists of an attribute-by-attribute comparison of the new problem description and each solved case. For the present invention several matching schemes will be implemented. During case creation, a knowledge engineer will be able to select the matching scheme most suitable to the problem at hand.
The following sections describe the matching schemes which are implemented in the present invention.
1) default_fuzzy_match
The default_fuzzy_match algorithm corresponds to the earlier implementation's handling of symbolic matching. A lookup table defined in the domain model is used to retrieve the similarity of any two values of the attribute.
As an example, consider the following attribute definition: attribute name: knife edge quality attribute values: new, sharp, corroded, dull similarity table: n ew sharp corroded dull n ew 1.0 0.8 0.3 0.0 sharp 1.0 0.6 0.0 corroded 1.0 0.5 dull 1.0
If a case contains the descriptor knife edge quality default _fuzzy_match sharp
The engine will compute the following: similarity (new) = 0.8 similarity (corroded) = 0.6 similarity (dull) = 0.0
The default_fuzzy_match algorithm may be used with nominal symbolic attributes. 2) custom_fuzzy_match The custom_fuzzy_match algorithm is essentially identical to default_fuzzy_match. However, a customized lookup table stored as part of the case is used instead of the default, global lookup table. Custom_fuzzy_match is provided for special cases in which the default_fuzzy_match table is not suitable. It is not anticipated that custom_fuzzy_match will be frequently used. Because of storage and performance penalties associated with custom_fuzzy_match, default_ fuzz y_ m atch should be used whenever possible. Custom_fuzzy_match is available for nominal symbolic attributes.
3) range
The range algorithm returns 1.0 if the attribute value falls on or within the specified limits, and 0.0 if it is outside the limits. For example, consider a case involving ice buildup on an aircraft wing. For arguments sake, assume icing only occurs between 10,000 and 15,000 feet altitude. The following representation would be used: altitude range 10000.0, 15000.0 The engine will compute the following: similarity (9999.0) = 0.0 similarity (10000.0) = 1.0 similarity (15000.0) = 1.0 similarity (15001.0) = 0.0 The range algorithm may be applied to ordered, integer, and floating point attributes.
4) less_than, fuzzy_less_than
The less_than algorithm returns 1.0 if the attribute value falls on or below the specified threshold, and 0.0 if it is above the threshold. For example, consider a case involving precipitation. For arguments sake, assume precipitation only occurs below 15,000 feet altitude. The following representation would be used: altitude less_than 15000.0
The engine will compute the following: similarity(WOOO.O) = 1.0 similarity (15000.0) = 1.0 similarity (15001.0) = 0.0 The less_than algorithm may be applied to ordered, integer, and floating point attributes.
The fuzzy_less_than algorithm is similar to the less_than algorithm. The only distinction is a gradual rather than abrupt transition from 1.0 to 0.0 in the similarity score at the threshold value.
5) greater_than, fuzzy_greater_than
The greater_than algorithm returns 1.0 if the attribute value falls on or above the specified threshold, and 0.0 if it is below the threshold. For example, consider a case involving a hydraulic seal leakage. The problem only occurs when the pressure differential across the seal is more than 4 atmospheres. The following representation would be used: pressure differential greater_than 4.0
The engine will compute the following: similarity '(3.999) = 0.0 similarity (4.0) = 1.0 similarity (5.0) = 1.0 The greater_than algorithm may be applied to ordered, integer, and floating point attributes. The fuzzy_greater_than algorithm is similar to the greater_than algorithm. The only distinction is a gradual rather than abrupt transition from 1.0 to 0.0 in the similarity score at the threshold value.
6) near_to
The near_to algorithm returns 1.0 if the attribute value in the problem description exactly matches the value in the stored case. The match level decreases to 0.0 as the values move apart from each other. The calculation is performed according to the following equation:
For example, consider a case involving propeller vibration due to a worn reduction gear. The vibration is most apparent when the propeller is operating at 2300 rpm. The following representation would be used: propeller rpm near_to 2300
The engine will compute the following: similarity(lOOO) = 0.00 similarity(2000) = 0.4 similarityQIOO) = 0.6 similarity(2300) = 1.0 similarity(2400) = 0.8 similarity (2700) = 0.2
The near_to algorithm may be applied to integer and floating point attributes.
7) range
The range algorithm returns 1.0 if the attribute value falls on or within the specified limits, and 0.0 if it is outside the limits. For example, consider a case involving ice buildup on an aircraft wing. For arguments sake, assume icing only occurs between 10,000 and 15,000 feet altitude. The following representation would be used: altitude range 10000.0, 15000.0
The engine will compute the following: similarity (9999.0) = 0.0 similarity (10000.0) = 1.0 similarity (15000.0) = 1.0 similarity (15001.0) = 0.0
The range algorithm may be applied to ordered, integer, and floating point attributes.
The similarity-computation schemes are classified according to the applicable attribute types. The following discussion deals with the computation of attribute similarity i.e.: the mapping from two values of a particular attribute to a number in the range 0.0 to 1.0. Throughout this discussion, the superscript "nc" on a variable refers to the variable's value in new problem case. The superscript "sc" refers to the variable's value in the solved case. The label "val" refers to value.
Figure 9 illustrates a similarity function definition using a quad- tuple representation, utilizing four parameters. Computation of numeric similarities is based on four parameters (hence, the name quad-tuple or set of four). These four parameters are a, b, c, and d. The parameters are defined relative to a reference value. This reference value is usually the value in the stored case. For most computations, the value of a, b , c, and d are specified in the units associated with the attribute. Comparison schemes based on the standard deviation information for an attribute may be implemented as well. Parameters for such schemes will be defined in standard deviation units.
The interpretation of the quad-tuple parameters is as follows: a. If the new value is lower than the reference value by this amount or more it is considered completely dissimilar. For example, consider a scenario where the temperature is lower by 15 degrees, resulting in significantly difference operating characteristics. b. If the new value is lower than the reference value by this amount or less it is considered essentially identical, (i.e., the decision maker is not concerned by the difference). For example, consider a scenario where the temperature is lower by 5 degrees, but the difference does not affect the outcome. c. If the new value is greater than the reference value by this amount or less it is considered essentially identical. d. If the new value is greater than the reference value by this amount or more it is considered completely dissimilar.
These parameters define the attribute similarity function about a particular value. A number of functional forms can be derived through different parameter settings (0 and ∞). An example of each functional form and its meaning follows.
The similarity computation is detailed in Table 3. Table 3: Numeric Similarity Computation Scheme
Figure imgf000035_0001
Figures 10 through 15 illustrate special case membership functions which can be derived by using different parameter settings. Each of these special cases correspond to a local similarity match algorithm as discussed earlier.
Special case I . b= 0 and c= 0. This corresponds to the near_to matching algorithm. See Figure 10.
Special case II. a= ∞, c= 0, and b = ∞. This corresponds to the fuzzy_less_than matching algorithm. See Figure 11.
Special case III. b= 0, c= ∞, and d= ∞. This corresponds to the fuzzy_greater_than matching algorithm. See Figure 12.
Special case IV. b= a, and d= c. This corresponds to the range matching algorithm. See Figure 13.
Special case V. b= a, and d= c= ∞. This corresponds to the greater_than matching algorithm.See Figure 14. Special case VI. b= a= ∞, and d = c. This corresponds to the less_than matching algorithm. See Figure 15.
The case specific scheme defines /redefines the global scheme with reference to the value used in a case. Although, this lends a great deal of flexibility for reasoning and for including context specific similarity assessment, the case specific specification should be avoided.
The same local similarity scheme can be used by symbolic ordinal or even numeric integer types to override other numeric computation schemes. The case specific similarity scheme is used to implement the custom_fuzzy_match similarity algorithm.
The must match scheme enforces a match with the attribute in which it is specified. That is, if the similarity with the specified attribute value is less than a preset threshold similarity value the overall similarity of the case (OSIM) is zero. The need for such a scheme frequently arises while representing applicability information in a case. For example, a D.C. motor case does not apply to an A.C. motor case. Therefore, the attribute "motor type =D.C." is specified with a must match similarity scheme.
The must match scheme is used locally in conjunction with any of the local similarity schemes. By default, the must match scheme is not enabled. The must match scheme includes a local similarity threshold value. The local similarity threshold value is specified by a linguistic label indicating the level of similarity. The labels used must be the same as those used in the similarity matrix. Consider the following example:
Attribute-value Local similarity Must match scheme scheme Motor type = D.C. {D.C., 1, A.C., 0} Must-m atch -s im ilar ity- threshold = Exact
In the example, the local similarity scheme says that the match of motor type with value D.C. is 1 and with value A.C. is 0. The must match scheme is interpreted as: the similarity of motor type must be at least exact for the case to be considered.
The overall similarity (OSIM) between a new problem case and a previously solved case is computed using the various matching functions of matcher module 36. Four exemplary matching functions are presented hereinbelow. It is assumed that no attribute weighting (i.e. domain knowledge or contextual knowledge) is provided with a new problem case description. This implies that the attributes of a new problem case are considered as equally weighted (e.g., 1). All the functions presented here are Nearest-Neighbour like because they deviate from the true nearest-neighbour function. The deviations include consideration of only a subset of all possible attributes for matching and the use of local importance of attributes. The local importance refers to the importance of an attribute in the context of a previously solved case (i.e., importance of an attribute is dependent on a previously solved case and recorded along with it). Nearest- neighbour uses global weights (i.e. importance of an attribute is the same across the whole case base). Prior art systems typically implement the true nearest-neighbour, and as a result do not consider local weights, nor can work with a subset of attributes.
Of the functions presented here, modified cosine matching function (OSIM 4) is considered to be the most sophisticated one. It performs significantly better than the nearest-neighbour when local importance along with subset of all possible attributes are used for matching. It also has the ability to consider contextual variation (i.e., difference between the importance of attributes of new problem case and a previously solved case). While this provides the ability to match contexts, it can make the matching oversensitive. The full contrast modified nearest-neighbour (OSIM 1) is a close second choice.
Figure imgf000039_0001
Table 4: OSIM
1. Nearest- The function operates
Neighbour (Full ∑ w, sim() over all relevant- contrast weighted attributes. The version) OSIM =. weights w of relevant
∑w attributes are assigned according to the CSC. For those attributes not in CSC they are equally weighted using a defined scheme.
2. Nearest- ∑w,sim() The function operates
Neighbour (Partial over CSC attributes contrast-CSC view OSIM = only. weighted version) ∑w
3. Nearest- ∑w,sim() The function operates
Neighbour (Partial over NC attributes contrast-NC view OSIM = only. weighted version) ∑w
4. Modified Cosine ∑ W ,NC W ,cscsim() This function Function operates over all
OSIM = relevant attributes, is ∑ (W,NC)2 X(wfCSC)2 capable of complete contrast and can take into account weights from NC and CSC Table 5: OSIM computing example
Figure imgf000040_0001
As will be apparent to those skilled in the art, various modifications and adaptations of the method and system described above are possible without departing from the present invention, the scope of which is defined in the appended claims.

Claims

WE CLAIM:
1. A method for assisting a user in solving a new problem case within a selected domain, comprising the steps of:
(a) providing a database comprising global domain knowledge relating to components of the selected domain, local domain knowledge, and a plurality of previously solved cases in the selected domain, each of the previously solved cases including a plurality of case attributes, said case attributes comprising case attribute names and known values associated therewith, said local domain knowledge comprising associations between the case attributes of a previously solved case;
(b) prompting the user to select a component of the domain and to select from the case attributes a set of attributes considered to be relevant to the new problem case and to provide current values for each of the new problem case attributes;
(c) searching the database of previously solved cases for candidate solved cases that include one or more of the new problem case attributes selected by the user and generating a list of said candidate solved cases;
(d) matching the candidate solved cases to the new problem case by comparing the current value for each of the new problem case attributes to the known value for the same case attribute in each of the candidate solved cases;
(e) ranking the candidate solved cases in order of relevance based upon the matching and presenting a list of ranked candidate solved cases in order of relevance based upon the ranking;
(f) generating additional questions based upon unanswered attributes of the candidate solved cases, the unanswered attributes being those case attributes for which values have not yet been provided by the user, and based upon the local domain knowledge, thereby assisting the user to select and provide values for the unanswered attributes; and
(g) repeating steps (b) through (f) inclusive until the user is satisfied with the list of candidate solved cases.
2. The method defined in claim 1, wherein the local domain knowledge comprises importance factors for the case attributes within a previously solved case, the importance factors being utilized in determining of which attributes questions should be first asked.
3. The method defined in claim 1, wherein the local domain knowledge comprises precedent constraints linking case attributes, within a previously solved case, the precedent constraints enabling questions related to the unanswered attributes to be generated only if the precedent constraints are satisfied.
4. The method defined in claim 1, wherein the local domain knowledge comprises match operators which enable values for case attributes relating to the new problem case to be matched with the known values for case attributes of previously solved cases.
5. The method as claimed in claim 2 wherein the matching step comprises the steps of:
(a) creating the list of candidate solved cases by reading each solved case and adding it to the list of solved candidate cases if any of the case attributes associated therewith match any of the new problem case attributes selected by the user for the new problem;
(b) generating an attribute similarity value for each case attribute of each solved case in the list of candidate solved cases; and
(c) generating an overall similarity value for each case in the list of candidate solved cases based upon the sum of the attribute similarity values.
6. The method defined in claim 5, wherein the step of generating additional questions is based upon:
a) the overall similarity of the new problem case to the candidate solved case;
b) the importance of the unanswered attribute in the candidate solved case and;
c) a vote value for the unanswered attribute in the new problem case created by summing the attribute importance by the overall similarity for each candidate solved case containing the unanswered attribute.
7. The method defined in claim 3 wherein the step of generating additional questions comprises asking questions related to the unanswered case attributes in descending order of overall similarity, attribute importance factor and vote value.
8. The method defined in claim 3, wherein the precedent constraints are satisfied when the user has provided values for the case attributes linked to the unanswered attribute, and the values exceed a predetermined threshold similarity value.
9. The method defined in claim 1, wherein the step of generating additional questions comprises the steps of:
(a) reading a candidate solved case from the list of ranked candidate solved cases;
(b) creating a list of case attributes from the candidate solved case, that have had all precedence requirements met before a question can be asked of the user;
(c) updating an initially empty list of unanswered case attributes having no value supplied by the user comprising the steps of:
(i) reading the list of case attributes created by step (b);
(ii) adding the case attribute to the list of attributes if it does not yet exist on the list;
(iii) increasing the vote value of the attribute identifier if it exists on the list;
(iv) repeating steps (i) to (iii) for each attribute. (d) repeating steps (a) to (c) for each case in the list of candidate solved cases;
(e) generating a sorted list from the list created by step (d); and
(f) asking the user questions in order of the attributes in the sorted list created by step (e).
10. The method defined in claim 1, wherein the database is a relational database.
11. The method defined in claim 10, wherein the database includes tables which define the hierarchical relationships between components of a piece of complex apparatus.
12. A computer system for assisting a user in solving a new problem case relating to a domain, comprising;
(a) storage means for storing a database comprising a plurality of solved cases, each of said solved cases comprising a plurality of case attributes, said case attributes comprising case attribute names and known values associated therewith, and local domain knowledge for the selected domain, the local domain knowledge comprising associations between the case attributes of a previously solved case;
(b) interface means for interfacing with the user, comprising output means for outputting to the user a list of case attributes of the solved cases and input means for enabling the user to select from the list of case attributes a set of attributes considered to be relevant to the new problem case, and to input current values for case attributes relating to a new problem case; and
(c) processing means coupled to the storage means and the interface means for processing the current values of the new problem case attributes, comprising:
(i) searching means for searching the database of previously solved cases for candidate solved cases;
(ii) matching means for matching the candidate solved cases to the new problem case by comparing the current values of the new problem case attributes with known values for the same case attributes for each of the candidate solved cases and determining the overall similarity of a new problem case with each candidate solved case; and
(iii) ranking means for ranking the candidate solved cases in order of relevance based upon the matching and creating a list of ranked candidate solved cases based upon said ranking.
13. The system defined in claim 12, wherein the processing means also comprises question generation means for generating additional questions based upon unanswered attributes in the candidate solved cases, the unanswered attributes being those case attributes for which values have not yet been provided by the user, to assist the user to enter additional current values for case attributes.
14. The system defined in claim 12, wherein the local domain knowledge comprises importance factors for the case attributes within a previously solved case, the importance factors being utilized in the determination of which attributes questions should first be asked.
15. The system defined in claim 12, wherein the local domain knowledge comprises precedent constraints linking case attributes, within a previously solved case, the precedent constraints enabling questions related to the unanswered attributes to be generated only if the precedent constraints are satisfied.
16. The system defined in claim 12, wherein the local domain knowledge comprises match operators which enable values for case attributes relating to the new problem case to be matched with the known values for case attributes of previously solved cases.
17. The system defined in claim 13, wherein the question generating means comprises :
a) means for determining the overall similarity of the new problem case to the candidate solved case;
b) means for determining the importance of the unanswered attribute in the candidate solved case; and
c) means for determining a vote value for the unanswered attribute in the new problem case created by summing the attribute importance by the overall similarity for each candidate solved case containing the unanswered attribute.
18. The system defined in claim 12, wherein the matching means comprises means for determining a similarity value for a given attribute of a given solution candidate case ranging from completely dissimilar to essentially identical based upon four parameters.
19. The system defined in claim 13, wherein the question generating means comprises means for selecting questions based upon the unanswered case attributes in descending order of overall similarity, attribute importance and vote value.
20. A method for assisting a user in solving a new problem case within a selected domain, comprising the steps of:
(a) providing a database comprising global domain knowledge relating to components of the selected domain, local domain knowledge, and a plurality of previously solved cases in the selected domain, each of the previously solved cases including a plurality of case attributes, said case attributes comprising case attribute names and known values associated therewith, said local domain knowledge comprising associations between the case attributes of a previously solved case;
(b) prompting the user to select a component of the domain and to select from the case attributes a set of attributes considered to be relevant to the new problem case and to provide current values for each of the new problem case attributes;
(c) searching the database of previously solved cases for candidate solved cases that include one or more of the new problem case attributes selected by the user and generating a list of said candidate solved cases; and
(d) matching the candidate solved cases to the new problem case by comparing the current value for each of the new problem case attributes to the known value for the same case attribute in each of the candidate solved cases, wherein the matching step comprises generating an attribute similarity value ranging from completely dissimilar to essentially identical based upon four parameters, the paremeters each defining a threshold of similarity or dissimilarity either less than or greater than a reference value.
21. The method defined in claim 20, wherein the similarity value is determined to be completely dissimilar if the value of the attribute in the new problem case is lower than the value of the corresponding attribute in the solved case by the first parameter or more.
22. The method defined in claim 20, wherein the similarity value is determined to be essentially identical if the value of attribute in the new problem case is lower than the value of the corresponding attribute in the solved case by the second parameter or less.
23. The method defined in claim 20, wherein the similarity value is determined to be essentially identical if the value of attribute in the new problem case is greater than the value of the corresponding attribute in the solved case by the third parameter or less.
24. The method defined in claim 20, wherein the similarity value is determined to be completely dissimilar if the value of attribute in the new problem case is greater than the value of the corresponding attribute in the solved case by the fourth parameter or more.
25. The method defined in claim 20, wherein the step of generating the attribute similarity value comprises the step of mapping the values of an attribute in the solved case and the corresponding attribute in the new problem case into a number in the range from 0.0 to 1.0.
26. The method defined in claim 20, wherein the matching steps also comprises the step of generating an overall similarity value for each of the candidate solved cases based upon the sum of the attribute similarity values.
27. The method defined in claim 26, wherein the step of generating an overall case similarity value comprises the steps of:
a) for each attribute in common between the previously solved case and the new problem case; summing the product of: the importance factor of the attribute in the new problem case, the importance factor of the attribute in the previously solved case, and the attribute similarity value;
b) summing the squares of the importance factor of each attribute in the new problem case;
c) summing the squares of the importance factor of each attribute in the previously solved case;
d) taking the square root of the results of steps b) and c); and
e) dividing the results of step a) by the results of step d)
28. The method defined in claim 1, wherein the case attributes comprise numeric attributes and symbolic attributes.
29. The method defined in claim 1, wherein the symbolic attributes comprise symbolic nominal attributes, and symbolic ordinal attributes.
30. The method defined in claim 28, wherein the matching step includes converting the values of the symbolic attributes into numbers using a look-up table.
PCT/CA1998/000306 1997-04-08 1998-04-03 Knowledge-based information retrieval system WO1998045775A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
AU69137/98A AU6913798A (en) 1997-04-08 1998-04-03 Knowledge-based information retrieval system
GB9922102A GB2338097B (en) 1997-04-08 1998-04-03 Knowledge-based information retrieval system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US08/835,558 1997-04-08
US08/835,558 US5822743A (en) 1997-04-08 1997-04-08 Knowledge-based information retrieval system

Publications (1)

Publication Number Publication Date
WO1998045775A1 true WO1998045775A1 (en) 1998-10-15

Family

ID=25269824

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CA1998/000306 WO1998045775A1 (en) 1997-04-08 1998-04-03 Knowledge-based information retrieval system

Country Status (5)

Country Link
US (1) US5822743A (en)
AU (1) AU6913798A (en)
CA (1) CA2206155C (en)
GB (1) GB2338097B (en)
WO (1) WO1998045775A1 (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1218831A1 (en) * 1999-09-22 2002-07-03 Infoglide Corporation System and method for performing similarity searching
EP1811520A2 (en) * 2006-01-17 2007-07-25 Omron Corporation Factor estimating device, method and program recording medium therefor
CN106663144A (en) * 2014-08-29 2017-05-10 皇家飞利浦有限公司 Method and apparatus for hierarchical data analysis based on mutual correlations
US10565539B2 (en) 2014-11-07 2020-02-18 International Business Machines Corporation Applying area of focus to workflow automation and measuring impact of shifting focus on metrics

Families Citing this family (226)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6571251B1 (en) * 1997-12-30 2003-05-27 International Business Machines Corporation Case-based reasoning system and method with a search engine that compares the input tokens with view tokens for matching cases within view
US6026393A (en) * 1998-03-31 2000-02-15 Casebank Technologies Inc. Configuration knowledge as an aid to case retrieval
US6263362B1 (en) * 1998-09-01 2001-07-17 Bigfix, Inc. Inspector for computed relevance messaging
US7197534B2 (en) 1998-09-01 2007-03-27 Big Fix, Inc. Method and apparatus for inspecting the properties of a computer
US8914507B2 (en) 1998-09-01 2014-12-16 International Business Machines Corporation Advice provided for offering highly targeted advice without compromising individual privacy
US6193518B1 (en) 1998-11-20 2001-02-27 Tina M. Nocera Method for developing answer-options to issue-questions relating to child-development
US6424969B1 (en) 1999-07-20 2002-07-23 Inmentia, Inc. System and method for organizing data
US6519578B1 (en) 1999-08-09 2003-02-11 Mindflow Technologies, Inc. System and method for processing knowledge items of a knowledge warehouse
US6519590B1 (en) * 1999-08-09 2003-02-11 Mindflow Technologies, Inc. System and method for performing a mindflow process using a mindflow document archive
US6629096B1 (en) * 1999-08-09 2003-09-30 Mindflow Technologies, Inc. System and method for performing a mindflow process
US6876991B1 (en) 1999-11-08 2005-04-05 Collaborative Decision Platforms, Llc. System, method and computer program product for a collaborative decision platform
US6571236B1 (en) * 2000-01-10 2003-05-27 General Electric Company Method and apparatus for problem diagnosis and solution
US6421571B1 (en) 2000-02-29 2002-07-16 Bently Nevada Corporation Industrial plant asset management system: apparatus and method
US7505921B1 (en) 2000-03-03 2009-03-17 Finali Corporation System and method for optimizing a product configuration
US8645137B2 (en) 2000-03-16 2014-02-04 Apple Inc. Fast, language-independent method for user authentication by voice
US20020147652A1 (en) * 2000-10-18 2002-10-10 Ahmed Gheith System and method for distruibuted client state management across a plurality of server computers
US7885820B1 (en) * 2000-07-19 2011-02-08 Convergys Cmg Utah, Inc. Expert system supported interactive product selection and recommendation
US7031951B2 (en) * 2000-07-19 2006-04-18 Convergys Information Management Group, Inc. Expert system adapted dedicated internet access guidance engine
US6745172B1 (en) 2000-07-19 2004-06-01 Whisperwire, Inc. Expert system adapted data network guidance engine
US7111010B2 (en) 2000-09-25 2006-09-19 Hon Hai Precision Industry, Ltd. Method and system for managing event attributes
WO2002027528A1 (en) * 2000-09-25 2002-04-04 Metaedge Corporation Method and system for managing event attributes
DE10107352B4 (en) * 2001-02-13 2007-11-22 T-Mobile Deutschland Gmbh Method and device for handling complaints on mobile phones
US6944619B2 (en) 2001-04-12 2005-09-13 Primentia, Inc. System and method for organizing data
US7401136B2 (en) 2001-07-27 2008-07-15 Dell Products L.P. Powertag: manufacturing and support system method and apparatus for multi-computer solutions
ITFI20010199A1 (en) 2001-10-22 2003-04-22 Riccardo Vieri SYSTEM AND METHOD TO TRANSFORM TEXTUAL COMMUNICATIONS INTO VOICE AND SEND THEM WITH AN INTERNET CONNECTION TO ANY TELEPHONE SYSTEM
GB2382678A (en) * 2001-11-28 2003-06-04 Symbio Ip Ltd a knowledge database
US7389280B2 (en) * 2002-02-22 2008-06-17 Lifecom, Inc. Computer-based intelligence method and apparatus for assessing selected subject-area problems and situations
US20070094197A1 (en) * 2002-02-22 2007-04-26 Datena Stephen J Medical diagnosis including graphical user input
JP4132962B2 (en) * 2002-05-16 2008-08-13 パイオニア株式会社 Interactive information providing apparatus, interactive information providing program, and storage medium storing the same
AU2003279115A1 (en) * 2002-10-03 2004-04-23 Whisperwire, Inc. System and method for bundling resources
US20040139107A1 (en) * 2002-12-31 2004-07-15 International Business Machines Corp. Dynamically updating a search engine's knowledge and process database by tracking and saving user interactions
US7216121B2 (en) * 2002-12-31 2007-05-08 International Business Machines Corporation Search engine facility with automated knowledge retrieval, generation and maintenance
US20040158561A1 (en) * 2003-02-04 2004-08-12 Gruenwald Bjorn J. System and method for translating languages using an intermediate content space
US7225176B2 (en) * 2003-03-26 2007-05-29 Casebank Technologies Inc. System and method for case-based reasoning
US7409593B2 (en) * 2003-06-30 2008-08-05 At&T Delaware Intellectual Property, Inc. Automated diagnosis for computer networks
US7324986B2 (en) * 2003-06-30 2008-01-29 At&T Delaware Intellectual Property, Inc. Automatically facilitated support for complex electronic services
US20050038697A1 (en) * 2003-06-30 2005-02-17 Aaron Jeffrey A. Automatically facilitated marketing and provision of electronic services
US7237266B2 (en) * 2003-06-30 2007-06-26 At&T Intellectual Property, Inc. Electronic vulnerability and reliability assessment
US20050108598A1 (en) * 2003-11-14 2005-05-19 Casebank Technologies Inc. Case-based reasoning system and method having fault isolation manual trigger cases
US20050193004A1 (en) * 2004-02-03 2005-09-01 Cafeo John A. Building a case base from log entries
US7392295B2 (en) 2004-02-19 2008-06-24 Microsoft Corporation Method and system for collecting information from computer systems based on a trusted relationship
US7584382B2 (en) * 2004-02-19 2009-09-01 Microsoft Corporation Method and system for troubleshooting a misconfiguration of a computer system based on configurations of other computer systems
US7389444B2 (en) * 2004-07-27 2008-06-17 Microsoft Corporation Method and system for troubleshooting a misconfiguration of a computer system based on product support services information
US8244689B2 (en) 2006-02-17 2012-08-14 Google Inc. Attribute entropy as a signal in object normalization
US7769579B2 (en) 2005-05-31 2010-08-03 Google Inc. Learning facts from semi-structured text
US8095392B1 (en) 2005-01-20 2012-01-10 Owen Daniel L System, method and computer program product for facilitating informed decisions relating to family risk
JP2006268405A (en) * 2005-03-24 2006-10-05 Hitachi Ltd Supporting device for creating customer value generation scenario, system and method
US7587387B2 (en) 2005-03-31 2009-09-08 Google Inc. User interface for facts query engine with snippets from information sources that include query terms and answer terms
US8682913B1 (en) 2005-03-31 2014-03-25 Google Inc. Corroborating facts extracted from multiple sources
US9208229B2 (en) 2005-03-31 2015-12-08 Google Inc. Anchor text summarization for corroboration
US7831545B1 (en) 2005-05-31 2010-11-09 Google Inc. Identifying the unifying subject of a set of facts
US7567976B1 (en) * 2005-05-31 2009-07-28 Google Inc. Merging objects in a facts database
US8996470B1 (en) 2005-05-31 2015-03-31 Google Inc. System for ensuring the internal consistency of a fact repository
US8677377B2 (en) 2005-09-08 2014-03-18 Apple Inc. Method and apparatus for building an intelligent automated assistant
US7492949B1 (en) 2005-09-20 2009-02-17 Patrick William Jamieson Process and system for the semantic selection of document templates
US7856100B2 (en) * 2005-12-19 2010-12-21 Microsoft Corporation Privacy-preserving data aggregation using homomorphic encryption
CA2530928A1 (en) * 2005-12-20 2007-06-20 Ibm Canada Limited - Ibm Canada Limitee Recommending solutions with an expert system
US8260785B2 (en) * 2006-02-17 2012-09-04 Google Inc. Automatic object reference identification and linking in a browseable fact repository
US7991797B2 (en) 2006-02-17 2011-08-02 Google Inc. ID persistence through normalization
US8700568B2 (en) 2006-02-17 2014-04-15 Google Inc. Entity normalization via name normalization
US9318108B2 (en) 2010-01-18 2016-04-19 Apple Inc. Intelligent automated assistant
US8122026B1 (en) 2006-10-20 2012-02-21 Google Inc. Finding and disambiguating references to entities on web pages
US8347202B1 (en) 2007-03-14 2013-01-01 Google Inc. Determining geographic locations for place names in a fact repository
US8977255B2 (en) 2007-04-03 2015-03-10 Apple Inc. Method and system for operating a multi-function portable electronic device using voice-activation
US8239350B1 (en) 2007-05-08 2012-08-07 Google Inc. Date ambiguity resolution
FI20075349L (en) * 2007-05-14 2008-11-15 Stiftelsen Arcada Method and arrangement in quality control
US7966291B1 (en) 2007-06-26 2011-06-21 Google Inc. Fact-based object merging
US7970766B1 (en) 2007-07-23 2011-06-28 Google Inc. Entity type assignment
US8738643B1 (en) 2007-08-02 2014-05-27 Google Inc. Learning synonymous object names from anchor texts
US7613694B2 (en) * 2007-08-24 2009-11-03 Omholt Ray E Computer-implemented method and apparatus for recording and organizing multiple responses to queries used to create a legacy profile in a manner that expands memory recall
US9053089B2 (en) 2007-10-02 2015-06-09 Apple Inc. Part-of-speech tagging using latent analogy
US8812435B1 (en) 2007-11-16 2014-08-19 Google Inc. Learning objects and facts from documents
US9330720B2 (en) 2008-01-03 2016-05-03 Apple Inc. Methods and apparatus for altering audio output signals
US8065143B2 (en) 2008-02-22 2011-11-22 Apple Inc. Providing text input using speech data and non-speech data
JP5182361B2 (en) * 2008-03-17 2013-04-17 富士通株式会社 Information acquisition support device
US8996376B2 (en) 2008-04-05 2015-03-31 Apple Inc. Intelligent text-to-speech conversion
US10496753B2 (en) 2010-01-18 2019-12-03 Apple Inc. Automatically adapting user interfaces for hands-free interaction
US8464150B2 (en) 2008-06-07 2013-06-11 Apple Inc. Automatic language identification for dynamic text processing
US20100030549A1 (en) 2008-07-31 2010-02-04 Lee Michael M Mobile device having human language translation capability with positional feedback
US8768702B2 (en) 2008-09-05 2014-07-01 Apple Inc. Multi-tiered voice feedback in an electronic device
US8898568B2 (en) 2008-09-09 2014-11-25 Apple Inc. Audio user interface
US20100063797A1 (en) * 2008-09-09 2010-03-11 Microsoft Corporation Discovering question and answer pairs
US8712776B2 (en) 2008-09-29 2014-04-29 Apple Inc. Systems and methods for selective text to speech synthesis
US8676904B2 (en) 2008-10-02 2014-03-18 Apple Inc. Electronic devices with voice command and contextual data processing capabilities
US9959870B2 (en) 2008-12-11 2018-05-01 Apple Inc. Speech recognition involving a mobile device
US8862252B2 (en) 2009-01-30 2014-10-14 Apple Inc. Audio user interface for displayless electronic device
US8380507B2 (en) 2009-03-09 2013-02-19 Apple Inc. Systems and methods for determining the language to use for speech generated by a text to speech engine
JP5325981B2 (en) * 2009-05-26 2013-10-23 株式会社日立製作所 Management server and management system
US9858925B2 (en) 2009-06-05 2018-01-02 Apple Inc. Using context information to facilitate processing of commands in a virtual assistant
US10241644B2 (en) 2011-06-03 2019-03-26 Apple Inc. Actionable reminder entries
US10540976B2 (en) 2009-06-05 2020-01-21 Apple Inc. Contextual voice commands
US20120311585A1 (en) 2011-06-03 2012-12-06 Apple Inc. Organizing task items that represent tasks to perform
US10241752B2 (en) 2011-09-30 2019-03-26 Apple Inc. Interface for a virtual digital assistant
US9431006B2 (en) 2009-07-02 2016-08-30 Apple Inc. Methods and apparatuses for automatic speech recognition
US8386410B2 (en) * 2009-07-22 2013-02-26 International Business Machines Corporation System and method for semantic information extraction framework for integrated systems management
US8682649B2 (en) 2009-11-12 2014-03-25 Apple Inc. Sentiment prediction from textual data
US8311838B2 (en) 2010-01-13 2012-11-13 Apple Inc. Devices and methods for identifying a prompt corresponding to a voice input in a sequence of prompts
US8381107B2 (en) 2010-01-13 2013-02-19 Apple Inc. Adaptive audio feedback system and method
US10276170B2 (en) 2010-01-18 2019-04-30 Apple Inc. Intelligent automated assistant
US10679605B2 (en) 2010-01-18 2020-06-09 Apple Inc. Hands-free list-reading by intelligent automated assistant
US10553209B2 (en) 2010-01-18 2020-02-04 Apple Inc. Systems and methods for hands-free notification summaries
US10705794B2 (en) 2010-01-18 2020-07-07 Apple Inc. Automatically adapting user interfaces for hands-free interaction
WO2011089450A2 (en) 2010-01-25 2011-07-28 Andrew Peter Nelson Jerram Apparatuses, methods and systems for a digital conversation management platform
US8682667B2 (en) 2010-02-25 2014-03-25 Apple Inc. User profiling for selecting user specific voice input processing information
US8713021B2 (en) 2010-07-07 2014-04-29 Apple Inc. Unsupervised document clustering using latent semantic density analysis
US8719006B2 (en) 2010-08-27 2014-05-06 Apple Inc. Combined statistical and rule-based part-of-speech tagging for text-to-speech synthesis
US8719014B2 (en) 2010-09-27 2014-05-06 Apple Inc. Electronic device with text error correction based on voice recognition data
US10762293B2 (en) 2010-12-22 2020-09-01 Apple Inc. Using parts-of-speech tagging and named entity recognition for spelling correction
US10515147B2 (en) 2010-12-22 2019-12-24 Apple Inc. Using statistical language models for contextual lookup
JP5648911B2 (en) * 2010-12-27 2015-01-07 インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Machines Corporation System, program and method
US8781836B2 (en) 2011-02-22 2014-07-15 Apple Inc. Hearing assistance system for providing consistent human speech
US9262612B2 (en) 2011-03-21 2016-02-16 Apple Inc. Device access using voice authentication
US10057736B2 (en) 2011-06-03 2018-08-21 Apple Inc. Active transport based notifications
US10672399B2 (en) 2011-06-03 2020-06-02 Apple Inc. Switching between text data and audio data based on a mapping
US8812294B2 (en) 2011-06-21 2014-08-19 Apple Inc. Translating phrases from one language into another using an order-based set of declarative rules
US8706472B2 (en) 2011-08-11 2014-04-22 Apple Inc. Method for disambiguating multiple readings in language conversion
US8994660B2 (en) 2011-08-29 2015-03-31 Apple Inc. Text correction processing
US8762156B2 (en) 2011-09-28 2014-06-24 Apple Inc. Speech recognition repair using contextual information
US10134385B2 (en) 2012-03-02 2018-11-20 Apple Inc. Systems and methods for name pronunciation
US9483461B2 (en) 2012-03-06 2016-11-01 Apple Inc. Handling speech synthesis of content for multiple languages
US9280610B2 (en) 2012-05-14 2016-03-08 Apple Inc. Crowd sourcing information to fulfill user requests
US10417037B2 (en) 2012-05-15 2019-09-17 Apple Inc. Systems and methods for integrating third party services with a digital assistant
US8775442B2 (en) 2012-05-15 2014-07-08 Apple Inc. Semantic search using a single-source semantic model
US10019994B2 (en) 2012-06-08 2018-07-10 Apple Inc. Systems and methods for recognizing textual identifiers within a plurality of words
US9721563B2 (en) 2012-06-08 2017-08-01 Apple Inc. Name recognition system
US9495129B2 (en) 2012-06-29 2016-11-15 Apple Inc. Device, method, and user interface for voice-activated navigation and browsing of a document
US9002769B2 (en) * 2012-07-03 2015-04-07 Siemens Aktiengesellschaft Method and system for supporting a clinical diagnosis
US9576574B2 (en) 2012-09-10 2017-02-21 Apple Inc. Context-sensitive handling of interruptions by intelligent digital assistant
US9547647B2 (en) 2012-09-19 2017-01-17 Apple Inc. Voice-based media searching
US8935167B2 (en) 2012-09-25 2015-01-13 Apple Inc. Exemplar-based latent perceptual modeling for automatic speech recognition
KR20230137475A (en) 2013-02-07 2023-10-04 애플 인크. Voice trigger for a digital assistant
US10642574B2 (en) 2013-03-14 2020-05-05 Apple Inc. Device, method, and graphical user interface for outputting captions
US10652394B2 (en) 2013-03-14 2020-05-12 Apple Inc. System and method for processing voicemail
US9368114B2 (en) 2013-03-14 2016-06-14 Apple Inc. Context-sensitive handling of interruptions
US10572476B2 (en) 2013-03-14 2020-02-25 Apple Inc. Refining a search based on schedule items
US9977779B2 (en) 2013-03-14 2018-05-22 Apple Inc. Automatic supplementation of word correction dictionaries
US9733821B2 (en) 2013-03-14 2017-08-15 Apple Inc. Voice control to diagnose inadvertent activation of accessibility features
US10748529B1 (en) 2013-03-15 2020-08-18 Apple Inc. Voice activated device for use with a voice-based digital assistant
US11151899B2 (en) 2013-03-15 2021-10-19 Apple Inc. User training by intelligent digital assistant
WO2014144579A1 (en) 2013-03-15 2014-09-18 Apple Inc. System and method for updating an adaptive speech recognition model
WO2014168730A2 (en) 2013-03-15 2014-10-16 Apple Inc. Context-sensitive handling of interruptions
US9922642B2 (en) 2013-03-15 2018-03-20 Apple Inc. Training an at least partial voice command system
WO2014197336A1 (en) 2013-06-07 2014-12-11 Apple Inc. System and method for detecting errors in interactions with a voice-based digital assistant
US9582608B2 (en) 2013-06-07 2017-02-28 Apple Inc. Unified ranking with entropy-weighted information for phrase-based semantic auto-completion
WO2014197334A2 (en) 2013-06-07 2014-12-11 Apple Inc. System and method for user-specified pronunciation of words for speech synthesis and recognition
WO2014197335A1 (en) 2013-06-08 2014-12-11 Apple Inc. Interpreting and acting upon commands that involve sharing information with remote devices
EP3937002A1 (en) 2013-06-09 2022-01-12 Apple Inc. Device, method, and graphical user interface for enabling conversation persistence across two or more instances of a digital assistant
US10176167B2 (en) 2013-06-09 2019-01-08 Apple Inc. System and method for inferring user intent from speech inputs
KR101809808B1 (en) 2013-06-13 2017-12-15 애플 인크. System and method for emergency calls initiated by voice command
JP6163266B2 (en) 2013-08-06 2017-07-12 アップル インコーポレイテッド Automatic activation of smart responses based on activation from remote devices
US20150127631A1 (en) * 2013-11-05 2015-05-07 International Business Machines Corporation Best available alternative dialog
US10296160B2 (en) 2013-12-06 2019-05-21 Apple Inc. Method for extracting salient dialog usage from live data
CA2909413C (en) * 2014-03-28 2019-06-18 Casebank Technologies Inc. Methods and systems for troubleshooting problems in complex systems using multiple knowledgebases
US9620105B2 (en) 2014-05-15 2017-04-11 Apple Inc. Analyzing audio input for efficient speech and music recognition
US10592095B2 (en) 2014-05-23 2020-03-17 Apple Inc. Instantaneous speaking of content on touch devices
US9502031B2 (en) 2014-05-27 2016-11-22 Apple Inc. Method for supporting dynamic grammars in WFST-based ASR
US9715875B2 (en) 2014-05-30 2017-07-25 Apple Inc. Reducing the need for manual start/end-pointing and trigger phrases
US10078631B2 (en) 2014-05-30 2018-09-18 Apple Inc. Entropy-guided text prediction using combined word and character n-gram language models
US9842101B2 (en) 2014-05-30 2017-12-12 Apple Inc. Predictive conversion of language input
US9734193B2 (en) 2014-05-30 2017-08-15 Apple Inc. Determining domain salience ranking from ambiguous words in natural speech
US9430463B2 (en) 2014-05-30 2016-08-30 Apple Inc. Exemplar-based natural language processing
US10289433B2 (en) 2014-05-30 2019-05-14 Apple Inc. Domain specific language for encoding assistant dialog
EP3480811A1 (en) 2014-05-30 2019-05-08 Apple Inc. Multi-command single utterance input method
US9785630B2 (en) 2014-05-30 2017-10-10 Apple Inc. Text prediction using combined word N-gram and unigram language models
US10170123B2 (en) 2014-05-30 2019-01-01 Apple Inc. Intelligent assistant for home automation
US9633004B2 (en) 2014-05-30 2017-04-25 Apple Inc. Better resolution when referencing to concepts
US9760559B2 (en) 2014-05-30 2017-09-12 Apple Inc. Predictive text input
US20150379419A1 (en) * 2014-06-26 2015-12-31 International Business Machines Corporation Ghost-pattern analyzer
US10659851B2 (en) 2014-06-30 2020-05-19 Apple Inc. Real-time digital assistant knowledge updates
US9338493B2 (en) 2014-06-30 2016-05-10 Apple Inc. Intelligent automated assistant for TV user interactions
US10446141B2 (en) 2014-08-28 2019-10-15 Apple Inc. Automatic speech recognition based on user feedback
US9818400B2 (en) 2014-09-11 2017-11-14 Apple Inc. Method and apparatus for discovering trending terms in speech requests
US10789041B2 (en) 2014-09-12 2020-09-29 Apple Inc. Dynamic thresholds for always listening speech trigger
US9886432B2 (en) 2014-09-30 2018-02-06 Apple Inc. Parsimonious handling of word inflection via categorical stem + suffix N-gram language models
US9668121B2 (en) 2014-09-30 2017-05-30 Apple Inc. Social reminders
US9646609B2 (en) 2014-09-30 2017-05-09 Apple Inc. Caching apparatus for serving phonetic pronunciations
US10127911B2 (en) 2014-09-30 2018-11-13 Apple Inc. Speaker identification and unsupervised speaker adaptation techniques
US10074360B2 (en) 2014-09-30 2018-09-11 Apple Inc. Providing an indication of the suitability of speech recognition
US10552013B2 (en) 2014-12-02 2020-02-04 Apple Inc. Data detection
US9711141B2 (en) 2014-12-09 2017-07-18 Apple Inc. Disambiguating heteronyms in speech synthesis
US9865280B2 (en) 2015-03-06 2018-01-09 Apple Inc. Structured dictation using intelligent automated assistants
US10567477B2 (en) 2015-03-08 2020-02-18 Apple Inc. Virtual assistant continuity
US9721566B2 (en) 2015-03-08 2017-08-01 Apple Inc. Competing devices responding to voice triggers
US9886953B2 (en) 2015-03-08 2018-02-06 Apple Inc. Virtual assistant activation
US9899019B2 (en) 2015-03-18 2018-02-20 Apple Inc. Systems and methods for structured stem and suffix language models
WO2016156433A1 (en) 2015-03-31 2016-10-06 British Telecommunications Public Limited Company Network operation
US9842105B2 (en) 2015-04-16 2017-12-12 Apple Inc. Parsimonious continuous-space phrase representations for natural language processing
US10083688B2 (en) 2015-05-27 2018-09-25 Apple Inc. Device voice control for selecting a displayed affordance
US11468101B2 (en) 2015-05-29 2022-10-11 Kuni Ahi LLC Context-rich key framework implementations for global concept management
US10127220B2 (en) 2015-06-04 2018-11-13 Apple Inc. Language identification from short strings
US10101822B2 (en) 2015-06-05 2018-10-16 Apple Inc. Language input correction
US10255907B2 (en) 2015-06-07 2019-04-09 Apple Inc. Automatic accent detection using acoustic models
US11025565B2 (en) 2015-06-07 2021-06-01 Apple Inc. Personalized prediction of responses for instant messaging
US10186254B2 (en) 2015-06-07 2019-01-22 Apple Inc. Context-based endpoint detection
GB2541034A (en) 2015-07-31 2017-02-08 British Telecomm Network operation
US10671428B2 (en) 2015-09-08 2020-06-02 Apple Inc. Distributed personal assistant
US10747498B2 (en) 2015-09-08 2020-08-18 Apple Inc. Zero latency digital assistant
US9697820B2 (en) 2015-09-24 2017-07-04 Apple Inc. Unit-selection text-to-speech synthesis using concatenation-sensitive neural networks
US11010550B2 (en) 2015-09-29 2021-05-18 Apple Inc. Unified language modeling framework for word prediction, auto-completion and auto-correction
US10366158B2 (en) 2015-09-29 2019-07-30 Apple Inc. Efficient word encoding for recurrent neural network language models
US11587559B2 (en) 2015-09-30 2023-02-21 Apple Inc. Intelligent device identification
US10691473B2 (en) 2015-11-06 2020-06-23 Apple Inc. Intelligent automated assistant in a messaging environment
US10049668B2 (en) 2015-12-02 2018-08-14 Apple Inc. Applying neural network language models to weighted finite state transducers for automatic speech recognition
US10223066B2 (en) 2015-12-23 2019-03-05 Apple Inc. Proactive assistance based on dialog communication between devices
US10446143B2 (en) 2016-03-14 2019-10-15 Apple Inc. Identification of voice inputs providing credentials
US9934775B2 (en) 2016-05-26 2018-04-03 Apple Inc. Unit-selection text-to-speech synthesis based on predicted concatenation parameters
US9972304B2 (en) 2016-06-03 2018-05-15 Apple Inc. Privacy preserving distributed evaluation framework for embedded personalized systems
US10249300B2 (en) 2016-06-06 2019-04-02 Apple Inc. Intelligent list reading
US10049663B2 (en) 2016-06-08 2018-08-14 Apple, Inc. Intelligent automated assistant for media exploration
DK179309B1 (en) 2016-06-09 2018-04-23 Apple Inc Intelligent automated assistant in a home environment
US10067938B2 (en) 2016-06-10 2018-09-04 Apple Inc. Multilingual word prediction
US10490187B2 (en) 2016-06-10 2019-11-26 Apple Inc. Digital assistant providing automated status report
US10192552B2 (en) 2016-06-10 2019-01-29 Apple Inc. Digital assistant providing whispered speech
US10509862B2 (en) 2016-06-10 2019-12-17 Apple Inc. Dynamic phrase expansion of language input
US10586535B2 (en) 2016-06-10 2020-03-10 Apple Inc. Intelligent digital assistant in a multi-tasking environment
DK179415B1 (en) 2016-06-11 2018-06-14 Apple Inc Intelligent device arbitration and control
DK201670540A1 (en) 2016-06-11 2018-01-08 Apple Inc Application integration with a digital assistant
DK179049B1 (en) 2016-06-11 2017-09-18 Apple Inc Data driven natural language event detection and classification
DK179343B1 (en) 2016-06-11 2018-05-14 Apple Inc Intelligent task discovery
US10593346B2 (en) 2016-12-22 2020-03-17 Apple Inc. Rank-reduced token representation for automatic speech recognition
DK179745B1 (en) 2017-05-12 2019-05-01 Apple Inc. SYNCHRONIZATION AND TASK DELEGATION OF A DIGITAL ASSISTANT
DK201770431A1 (en) 2017-05-15 2018-12-20 Apple Inc. Optimizing dialogue policy decisions for digital assistants using implicit feedback
US20180330802A1 (en) * 2017-05-15 2018-11-15 Koninklijke Philips N.V. Adaptive patient questionnaire generation system and method
CN111259659B (en) * 2020-01-14 2023-07-04 北京百度网讯科技有限公司 Information processing method and device
US11763182B1 (en) * 2020-05-07 2023-09-19 Jared Anders Newcombe Software facilitating decision making method
CN112861276B (en) * 2021-01-12 2022-11-08 北京理工大学 Blast furnace burden surface optimization method based on data and knowledge dual drive

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5319739A (en) * 1990-11-22 1994-06-07 Hitachi, Ltd. Method for retrieving case for problem and inferring solution of the problem and system thereof
EP0694836A2 (en) * 1994-07-28 1996-01-31 Hitachi Europe Limited A case base system

Family Cites Families (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4658370A (en) * 1984-06-07 1987-04-14 Teknowledge, Inc. Knowledge engineering tool
US4697243A (en) * 1985-07-25 1987-09-29 Westinghouse Electric Corp. Methods of servicing an elevator system
US4763277A (en) * 1986-01-17 1988-08-09 International Business Machines Corporation Method for obtaining information in an expert system
US4964125A (en) * 1988-08-19 1990-10-16 Hughes Aircraft Company Method and apparatus for diagnosing faults
US4985857A (en) * 1988-08-19 1991-01-15 General Motors Corporation Method and apparatus for diagnosing machines
US5282265A (en) * 1988-10-04 1994-01-25 Canon Kabushiki Kaisha Knowledge information processing system
US5107497A (en) * 1989-07-28 1992-04-21 At&T Bell Laboratories Technique for producing an expert system for system fault diagnosis
JP3224226B2 (en) * 1989-09-22 2001-10-29 株式会社リコー Fault diagnosis expert system
US5224206A (en) * 1989-12-01 1993-06-29 Digital Equipment Corporation System and method for retrieving justifiably relevant cases from a case library
JP3266246B2 (en) * 1990-06-15 2002-03-18 インターナシヨナル・ビジネス・マシーンズ・コーポレーシヨン Natural language analysis apparatus and method, and knowledge base construction method for natural language analysis
US5586218A (en) * 1991-03-04 1996-12-17 Inference Corporation Autonomous learning and reasoning agent
EP0575473B1 (en) * 1991-03-04 2000-04-26 Inference Corporation Case-based reasoning system
US5317677A (en) * 1992-04-16 1994-05-31 Hughes Aircraft Company Matching technique for context sensitive rule application
US5402524A (en) * 1992-12-22 1995-03-28 Mitsubishi Denki Kabushiki Kaisha Case-based knowledge source for artificial intelligence software shell
JP3067966B2 (en) * 1993-12-06 2000-07-24 松下電器産業株式会社 Apparatus and method for retrieving image parts
US5644686A (en) * 1994-04-29 1997-07-01 International Business Machines Corporation Expert system and method employing hierarchical knowledge base, and interactive multimedia/hypermedia applications
US5715468A (en) * 1994-09-30 1998-02-03 Budzinski; Robert Lucius Memory system for storing and retrieving experience and knowledge with natural language
US5717835A (en) * 1995-01-11 1998-02-10 International Business Machines Corporation Simple approach to case-based reasoning for data navigation tasks
JPH08194618A (en) * 1995-01-18 1996-07-30 Kobe Steel Ltd Case base inferring device

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5319739A (en) * 1990-11-22 1994-06-07 Hitachi, Ltd. Method for retrieving case for problem and inferring solution of the problem and system thereof
EP0694836A2 (en) * 1994-07-28 1996-01-31 Hitachi Europe Limited A case base system

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1218831A1 (en) * 1999-09-22 2002-07-03 Infoglide Corporation System and method for performing similarity searching
EP1218831A4 (en) * 1999-09-22 2007-09-19 Infoglide Corp System and method for performing similarity searching
EP1811520A2 (en) * 2006-01-17 2007-07-25 Omron Corporation Factor estimating device, method and program recording medium therefor
EP1811520A3 (en) * 2006-01-17 2008-12-31 Omron Corporation Factor estimating device, method and program recording medium therefor
CN106663144A (en) * 2014-08-29 2017-05-10 皇家飞利浦有限公司 Method and apparatus for hierarchical data analysis based on mutual correlations
US10565539B2 (en) 2014-11-07 2020-02-18 International Business Machines Corporation Applying area of focus to workflow automation and measuring impact of shifting focus on metrics

Also Published As

Publication number Publication date
GB9922102D0 (en) 1999-11-17
AU6913798A (en) 1998-10-30
CA2206155A1 (en) 1998-10-08
CA2206155C (en) 2000-03-21
US5822743A (en) 1998-10-13
GB2338097B (en) 2002-03-06
GB2338097A (en) 1999-12-08

Similar Documents

Publication Publication Date Title
US5822743A (en) Knowledge-based information retrieval system
US7689544B2 (en) Automatic indexing of digital image archives for content-based, context-sensitive searching
Yang et al. Using corpus statistics to remove redundant words in text categorization
US7734556B2 (en) Method and system for discovering knowledge from text documents using associating between concepts and sub-concepts
Li et al. An information filtering model on the Web and its application in JobAgent
Chajewska et al. Utility Elicitation as a Classification Problem.
US20090106225A1 (en) Identification of medical practitioners who emphasize specific medical conditions or medical procedures in their practice
Chen et al. Methods for processing and prioritizing customer demands in variant product design
Sadiq et al. Hybrid intelligent technique for text categorization
Bordogna et al. Modeling vagueness in information retrieval
Hu Personalized web search by using learned user profiles in re-ranking
Dubitzky et al. An advanced case-knowledge architecture based on fuzzy objects
Radwan et al. Thyroid diagnosis based technique on rough sets with modified similarity relation
Lieber Strong, fuzzy and smooth hierarchical classification for case-based problem solving
Reza Montazemi et al. A framework for retrieval in case-based reasoning systems
Ho et al. Documents clustering using tolerance rough set model and its application to information retrieval
Remadi et al. The triangular intuitionistic fuzzy extension of the CODAS method for solving multi-criteria group decision making
Brini et al. Relevance feedback: introduction of partial assessments for query expansion.
Voglozin et al. Querying a summary of database
Duval Abduction for explanation-based learning
WO2004006124A2 (en) Text-representation, text-matching and text-classification code, system and method
Vale et al. Improving text retrieval in medical collections through automatic categorization
Falchuk et al. Location-Based Services
De Tré et al. Conjunctive aggregation of extended possibilistic truth values and flexible database querying
Hsieh Rule extraction with rough-fuzzy hybridization method

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AL AM AT AU AZ BA BB BG BR BY CA CH CN CU CZ DE DK EE ES FI GB GE GH GM GW HU ID IL IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT UA UG US UZ VN YU ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH GM KE LS MW SD SZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN ML MR NE SN TD TG

DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
121 Ep: the epo has been informed by wipo that ep was designated in this application
ENP Entry into the national phase

Ref country code: GB

Ref document number: 9922102

Kind code of ref document: A

Format of ref document f/p: F

REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

NENP Non-entry into the national phase

Ref country code: JP

Ref document number: 1998542177

Format of ref document f/p: F

122 Ep: pct application non-entry in european phase
NENP Non-entry into the national phase

Ref country code: CA