US 20060112045 A1
Systems and methods are provided for assisted decision-making. A knowledge base comprises a plurality of executable stories. A given story comprises a belief network containing at least one associated hypothesis and evidence associated with the at least one hypothesis. The system further comprises at least one decision algorithm having respective associated input formats. A translation interface is operative to convert a given story from the knowledge base into a decision network having a format associated with the at least one decision algorithm.
1. An assisted decision-making system comprising:
a knowledge base comprising a plurality of executable stories, a given executable story comprising a belief network containing at least one associated hypothesis and evidence associated with the at least one hypothesis;
at least one decision algorithm having respective associated input formats; and
a translation interface operative to convert a given story from the knowledge base into a decision network having a format associated with the at least one decision algorithm.
2. The system of
3. The system of
4. The system of
5. The system of
6. The system of
7. The system of
8. The system of
9. The system of
10. The system of
11. A method for assisting decision making comprising:
storing a plurality of executable stories in a knowledge base;
converting a given story from the plurality of executable stories into a decision network having a format associated with a decision algorithm associated with the knowledge base;
processing the decision network at the decision algorithm to produce a revised decision network; and
converting the revised decision network into a revised executable story.
12. The method of
13. The method of
14. The method of
15. The method of
16. The method of
17. An assisted decision-making system comprising:
means for storing a plurality of executable stories, a given story comprising a belief network containing at least one associated hypothesis and evidence associated with the at least one hypothesis;
means for processing a decision network, the means for processing having an associated input format; and
means for translating a given story from the means for storing into a decision network having the input format associated with the means for processing.
18. The system of
19. The system of
20. The system of
21. The system of
22. The system of
23. The system of
The present invention relates to artificial intelligence and, more particularly, to systems and methods for decision making utilizing a knowledge base comprising executable stories.
There are different approaches for organizing information content extracted from text, such as decision graphs or data tables, namely, a knowledge-based approach and a data-based approach. Using the knowledge-based approach, a person (known as a knowledge engineer) interviews an expert in a given field to obtain knowledge about the given field. The knowledge engineer and expert first determine the distinctions of the subject matter that are important for decision making in the field of the expert. These distinctions correspond to questions about the variables in the domain of interest, referred to as the hypotheses. For example, if a decision graph is to be used to predict the age of a customer based on the products that customer bought in a store, there would be a variable for “age” and a variable for all relevant products. The knowledge engineer and the expert next determine the structure of the decision graph and the corresponding parameter values that quantify the associated uncertainty.
In the database approach, the knowledge engineer and the expert first determine the variables of the domain. Next, data is accumulated for those variables, and an algorithm is applied that creates one or more decision graphs from this data. The accumulated data comes from real world instances of the domain or hypothesis. That is, real world instances of decision making in a given field. For some decision-making applications, however, it can be difficult in practice to find sufficient applicable data to construct a viable decision network.
Decision makers often find it difficult to mentally combine evidence since the human tendency is to postpone risky decisions when data is incomplete, jump to conclusions, or refuse to consider conflicting data. Those versed in classical (frequentist) statistics realize that in situations where evidence is sparse, the use of a decision network is desirable. Unfortunately, traditional means of generating a decision network are labor intensive, requiring many hours of labor from an analyst knowledgeable about the desired application.
In accordance with one aspect of the present invention, an assisted decision-making system is provided. A knowledge base comprises a plurality of executable stories. A given story comprises a belief network containing at least one associated hypothesis and evidence associated with the at least one hypothesis. Each hypothesis can have labeled characteristics and values of characteristics. The system further comprises at least one decision algorithm having respective associated input formats. A translation interface is operative to convert a given story from the knowledge base into a decision network having a format associated with the at least one decision algorithm.
In accordance with another aspect of the present invention, a method for assisting decision making is provided. A plurality of executable stories are stored in a knowledge base. A story from the plurality of executable stories is converted into a decision network in a format associated with a decision algorithm associated with the knowledge base. The decision network is processed at the decision algorithm to produce a revised decision network. The revised decision network is converted back into an executable story.
The present invention relates to systems and methods for assisted decision making utilizing a knowledge base that comprises a plurality of stories, where a story is an executable belief network augmented by one or more characteristics of the hypotheses comprising the belief network, the evidence, and the content from which the evidence was extracted. By constructing the knowledge base from executable elements, the stories can be manipulated and updated in real time with new evidence. For example, a union or intersection between two stories or portions of stories can be formed, with appropriate belief values calculated from the known values, without the need for human intervention. Further, the knowledge base can be meaningfully shared between decision makers who may utilize different bodies of evidence.
In certain aspects of the invention, the stories comprising the knowledge base can include Dempster-Shafer belief networks. Dempster-Shafer belief network include node parameters that conform to the Dempster-Shafer combination rule, which is based on an evidential interval: the sum of a belief value, a disbelief value, and an unknown value is equal to one. The Dempster-Shafer Combination Rule for fusion of evidence provides for nodes in a network represented as evidential intervals with values from the set of real numbers (0<=n<=1). Three parameters specify each node: “belief” (B), “unknown” (U) and “disbelief” (D). The unknown parameter is computed as: U=1−B−D. The Dempster-Shafer Combination Rule is symmetric, bounded, commutative, and associative.
It will be appreciated that the stories in the knowledge base incorporate executable belief networks and a body of evidence for evaluating their associated belief network. By executable, it is meant that the stories in the knowledge base are objects capable of independently updating their internal values (e.g., in accordance with the Dempster-Shafer combinational rule or Bayesian theory) to achieve a mathematically correct state given an external perturbation. The executable nature of these stories allows an analyst a great degree of freedom in forecasting the response of the various decision algorithms 14 and 16 to new evidence or changed, as a change in the central knowledge base can be quickly recognized and incorporated into the analysis of the decision algorithms 14 and 16. Further, the executable nature of the stories allows a given story to be evaluated independently of the knowledge base, allowing the stories to be easily transferred between multiple systems or analysts.
The plurality of decision algorithms 14 and 16 have respective input formats for importing decision networks associated with the assisted decision making system 10. For example, a case based reasoning algorithm can process decision networks in tabular form, comprising a table of criteria and scores relating to a plurality of options considered by the algorithm. The knowledge base 12 and the decision algorithms are connected via the translation interface 18 that is operative to convert an executable story into a decision network in one or more of the input formats associated with the plurality of decision algorithms. For example, where one of the decision algorithms (e.g., 14) is a case based reasoning algorithm, a story or portion of a story from the knowledge base 12 can be converted into a tabular form consistent with the case based reasoning algorithm at the translation interface 18 and then provided to the algorithm for analysis. Accordingly, all of the decision algorithms can draw on the central knowledge base while receiving data in their associated formats.
Similarly, the translation interface 18 can convert the decision outputs from the plurality of decision making algorithms 14 and 16 into stories. For example, the translation interface 18 can convert a tabular output from a case based reasoning algorithm by creating hypotheses from the options in the table rows from and supporting them with evidence provided within the columns of the table, such as the criteria utilized by the system. The scores determined by the algorithm, along with other information within the table, can be used to determine weight values for the evidence provided. Accordingly, the stories stored in the knowledge base can comprise stories submitted to the plurality of decision algorithms 14 and 16, as well as evidence added to the knowledge base via a user interface and an associated information extraction component.
The links represent multipliers or weights of a given parameter on a lower node. Link values can be constant, or computed by an algorithm. For example, the belief of node N7 of the first intermediate layer 54 depends on the belief of nodes N1, N2, and N3, each multiplied by its respective link value L1, L2, and L3. Additionally, the disbelief of node N7 of the first intermediate layer 54 depends on the disbelief of nodes N1, N2, and N3, each multiplied by its respective link value L1, L2, and L3. The unknown is computed based on the Dempster-Shafer combination rule. The belief and disbelief of node N7 then propagate to N11 through link L11, which is combined with the belief and disbelief of N18 multiplied by link L12 and the belief and disbelief of node N9 multiplied by link L14. The belief and disbelief of node N11 then propagate to node N14 through link L18 which is combined with the belief and disbelief of N13 multiplied by link L20. The unknowns of each row can be evaluated using the Dempster-Shafer combination rule. Similar propagation occurs to provide the beliefs, the disbeliefs, and unknowns of the node N15.
In accordance with an aspect of the present invention, the knowledge base 152 is comprised of a plurality of stories, where each story comprises an executable belief network comprising at least one hypothesis, evidence supporting the at least one hypothesis, and a reference (e.g., a pointer) to the context from which the evidence was gathered. Each story is executable, such that it can produce mathematically consistent results in response to any change in its associated evidence, belief values, or weights. Accordingly, the stories can be updated and propagated to multiple decision algorithms in real time, allowing for a flexible exchange between a large number of decision algorithms or analysts.
In the illustrated example, evidence and stories can be input into the knowledge base in a number of ways. For example, an information extraction component 154 can be used to reduce an evidence source, such as a text document or a transcripted conversation, into a desired evidence format. This evidence can be linked with existing stories in the knowledge base or new stories can be assembled in response to the evidence. Stories and portions of stories can also be provided to the knowledge base or amended within the knowledge base using other decision algorithms 156 and 158. For example, the illustrated example includes a case-base reasoning algorithm 156 that utilizes existing stories in the knowledge base to improve or repair a decision network being constructed by the system 150. Similarly, a data mining algorithm 158 can determine additional hypotheses and linkages among available hypotheses in the knowledge base 152 to refine a decision network being constructed by the system.
The information extraction component 154 breaks down the text segments into individual words or phrases, interprets the context and meaning of the various words or phrases, and uses the extracted information to generate a template representing the text segment. For example, the information extraction component 154 can look for details relating to an event described in the document, such as the nature of the event, the cause or motivation for the event, the mechanism of the event, the identity of an actor, the location of the event, the time or date of the event, and the magnitude of the event. Each of these details can be added to a template related to the text segment. In accordance with one aspect of the invention, the information extraction component 154 can look for hedge words (e.g., maybe, probably, certainly, never) within the text segment. The information extraction component 154 can use a co-referencing routine to determine what nouns relate to a given hedge word, and use this information to determine the weight of the evidence associated with the template, in the form of belief values and disbelief values.
To provide a greatly simplified example, the information extraction component 154 might receive a statement from a bank teller that they are certain that Mr. Brown has made a deposit of ten-thousand dollars to a corporate account via a personal check at a bank in downtown Atlanta. The information extraction component 154 would locate the nouns within the sentence as well as words such as “via,” “at,” and “certain” to determine the relationships between the various nouns and the location of certain information. Thus, the question of location can be answered with the noun or string of nouns following “at” (e.g., bank in downtown Atlanta). The mechanism of the event can be determined by the nouns following “via” (e.g., personal check). The magnitude of the event can be determined by finding the numbers (e.g., ten-thousand), and other details can be provided by classifying the remaining nouns (e.g., Mr. Brown is likely the actor; the event is likely a deposit, etc.). The word “certain,” once it is verified that it is referring the deposit, can be used to assign a large belief value to the event.
During operation, objects of evidence can be provided to one or more evidence classifiers 160. The evidence classifiers 160 assign the evidence to associated hypotheses according to the evidence content. It will be appreciated that the evidence classifiers 160 can assign the templates to one or more existing hypotheses in the knowledge base 152 or generate a new hypothesis (e.g., from a large independent ontology or the Internet). The hypotheses within each network can be derived from previously generated networks, new hypotheses added to accommodate additional evidence, and a priori knowledge of the problem added by an analyst through a user interface 162. In an exemplary embodiment, the evidence classifiers 160 can include a rule-based classifier that classifies the templates according to a set of user defined rules. For example, rules can be defined relating to the fields within the template or the source of the data. Other classifiers can include, for example, supervised and unsupervised neural network classifiers, semantic network classifiers, statistical classifiers, and other classifier models. These classifiers can be orchestrated to increase the efficiency of the classification. For example, the rule-based classifier can be applied first, and if a rule is not actuated, the statistical classifier can be used. If a pre-specified probability threshold is not reached at the statistical classifier, the semantic distance classifier can be applied and the results shown to the user for validation.
Once the templates have been assigned to appropriate hypotheses, the templates are passed to a link analysis component 164. The link analysis component 164 determines the relatedness of two hypotheses according to relationships in the evidence associated with the templates. One factor utilized by the link analysis component 164 is the co-occurrence of various key words and phrases in the evidence templates assigned to two hypotheses. The frequency of co-occurrence of evidence among the various hypotheses can be used to calculate initial strength values, or weights, to the links between hypotheses. Link strength indicates the degree of influence a given hypothesis and its associated belief and disbelief values have on a linked hypothesis.
The link strengths and hypotheses are then provided to a fusion engine 166. The fusion engine 166 arranges the hypotheses according to a predefined structure to establish a hierarchy among the hypotheses. For example, the predefined structure can include a number of categories, representing layers within the hierarchy, with the proper category for a given hypothesis being determined when it is entered into the knowledge base 152. Once the hypotheses have been categorized, they can be linked according to the determined link strengths to form a preliminary belief network. In the illustrated example, a Dempster-Shafer belief network is utilized, but it will be appreciated that other belief network architectures (e.g., Bayesian) can be used as well. This preliminary network can be reviewed by a human analyst through the user interface 162 to ensure that the hypotheses are organized and linked correctly. The network is then fused with its associated evidence and the context from which the evidence was drawn to form an executable story. The executable story can mathematically reconcile its associated belief, disbelief, and unknown values according to the Dempster-Shafer combinational rule.
In accordance with an aspect of the present invention, the fusion engine 166 can also provide internal management of the stories stored within the knowledge base. For example, the fusion engine 166 can also be used to add, remove, or modify hypothesis and evidence in an existing story in the knowledge base in response to new evidence or user input. Similarly, the fusion engine 166 can be rank stories by average belief or disbelief in the hypotheses that comprise the underlying belief networks and maintain them within the knowledge base in ranked form.
Executable stories generated at the fusion engine 166 are stored in the knowledge base 152, where they can be accessed and modified by other decision algorithms 156 and 158 associated with the system. The executable stories can be provided to the decision algorithms through respective interface modules 168 and 170 to allow the decision algorithms to further refine the stories. For example, the stories in the knowledge base 152 can be searched by a data mining algorithm 158 through its associated interface module 170. The interface module 170 transmits the stories to the data mining algorithm 158 in an appropriate input format, such as a directed graph. It will be appreciated that the desired input form will vary with the specifics of the data mining algorithm. The data mining algorithm 158 can examine existing hypothesis within stories stored in the knowledge base 152 to determine appropriate one or more appropriate hypotheses for a body of evidence. These hypotheses can then be used to generate a new story or supplement an existing story stored in the knowledge base.
Similarly, a story of interest can be provided to a case based reasoning component 156 through its associated interface module 168 when it meets certain selection criteria. The interface module 168 associated with the case based reasoning component can translate the executable stories into a decision network in tabular form appropriate for input into the case based reasoning component 156. To refine or correct the story of interest, the case based reasoning component 156 can access successful stories located in the knowledge base in the tabular format through the interface module 168 and implement portions of the successful stories into the decision network as new network fragments. For example, when a series of evidence templates that are not associated with a known hypothesis are inputted to the system 150, the case based reasoning component 156 can determine if similar evidence is present within a story stored with the knowledge base. If so, one or more hypotheses from the stored story can be adopted (e.g., by computing the union the story of interest and a portion of the stored story) to produce a refined instantiation of the story of interest. The decision network can then be translated back into an executable story for storage in the knowledge base.
In view of the foregoing structural and functional features described above, methodology in accordance with various aspects of the present invention will be better appreciated with reference to
At 256, the decision algorithm processes the decision network to produce a revised network. For example, a case based reasoning algorithm can review past successful stories to add or modify a portion of the decision network. Similarly, a data mining algorithm can draw upon stories in the knowledge base to determine additional hypotheses, links, or evidence for a given decision network. Once the decision network has been revised, it can be translated back into an executable story at step 258. The executable story can be utilized for decision making or stored in the knowledge base for later reference.
The computer system 300 includes a processor 302 and a system memory 304. A system bus 306 couples various system components, including the system memory 304 to the processor 302. Dual microprocessors and other multiprocessor architectures can also be utilized as the processor 302. The system bus 306 can be implemented as any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, and a local bus using any of a variety of bus architectures. The system memory 304 includes read only memory (ROM) 308 and random access memory (RAM) 310. A basic input/output system (BIOS) 312 can reside in the ROM 308, generally containing the basic routines that help to transfer information between elements within the computer system 300, such as a reset or power-up.
The computer system 300 can include a hard disk drive 314, a magnetic disk drive 316, e.g., to read from or write to a removable disk 318, and an optical disk drive 320, e.g., for reading a CD-ROM or DVD disk 322 or to read from or write to other optical media. The hard disk drive 314, magnetic disk drive 316, and optical disk drive 320 are connected to the system bus 306 by a hard disk drive interface 324, a magnetic disk drive interface 326, and an optical drive interface 334, respectively. The drives and their associated computer-readable media provide nonvolatile storage of data, data structures, and computer-executable instructions for the computer system 300. Although the description of computer-readable media above refers to a hard disk, a removable magnetic disk and a CD, other types of media which are readable by a computer, may also be used. For example, computer executable instructions for implementing systems and methods described herein may also be stored in magnetic cassettes, flash memory cards, digital video disks and the like.
A number of program modules may also be stored in one or more of the drives as well as in the RAM 310, including an operating system 330, one or more application programs 332, other program modules 334, and program data 336.
A user may enter commands and information into the computer system 300 through user input device 340, such as a keyboard, a pointing device (e.g., a mouse). Other input devices may include a microphone, a joystick, a game pad, a scanner, a touch screen, or the like. These and other input devices are often connected to the processor 302 through a corresponding interface or bus 342 that is coupled to the system bus 306. Such input devices can alternatively be connected to the system bus 306 by other interfaces, such as a parallel port, a serial port or a universal serial bus (USB). One or more output device(s) 344, such as a visual display device or printer, can also be connected to the system bus 306 via an interface or adapter 346.
The computer system 300 may operate in a networked environment using logical connections 348 to one or more remote computers 350. The remote computer 348 may be a workstation, a computer system, a router, a peer device or other common network node, and typically includes many or all of the elements described relative to the computer system 300. The logical connections 348 can include a local area network (LAN) and a wide area network (WAN).
When used in a LAN networking environment, the computer system 300 can be connected to a local network through a network interface 352. When used in a WAN networking environment, the computer system 300 can include a modem (not shown), or can be connected to a communications server via a LAN. In a networked environment, application programs 332 and program data 336 depicted relative to the computer system 300, or portions thereof, may be stored in memory 354 of the remote computer 350.
What has been described above includes exemplary implementations of the present invention. It is, of course, not possible to describe every conceivable combination of components or methodologies for purposes of describing the present invention, but one of ordinary skill in the art will recognize that many further combinations and permutations of the present invention are possible. For example, while a number of examples herein have discussed the construction of Dempster-Shafer belief networks, it will be appreciated that other decision networks, such as Bayesian belief networks and Markov networks, can be constructed from text in accordance with the present invention. Accordingly, the present invention is intended to embrace all such alterations, modifications, and variations that fall within the spirit and scope of the appended claims.