WO2006076398A2 - Predictive analytic method and apparatus - Google Patents

Predictive analytic method and apparatus Download PDF

Info

Publication number
WO2006076398A2
WO2006076398A2 PCT/US2006/000908 US2006000908W WO2006076398A2 WO 2006076398 A2 WO2006076398 A2 WO 2006076398A2 US 2006000908 W US2006000908 W US 2006000908W WO 2006076398 A2 WO2006076398 A2 WO 2006076398A2
Authority
WO
WIPO (PCT)
Prior art keywords
project
pof
ontology
data
message
Prior art date
Application number
PCT/US2006/000908
Other languages
French (fr)
Other versions
WO2006076398B1 (en
WO2006076398A3 (en
Inventor
Douglas Clark
Brian Pieslak
Brian Gipson
Zachary Walton
Original Assignee
Metier Ltd
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 Metier Ltd filed Critical Metier Ltd
Publication of WO2006076398A2 publication Critical patent/WO2006076398A2/en
Publication of WO2006076398A3 publication Critical patent/WO2006076398A3/en
Publication of WO2006076398B1 publication Critical patent/WO2006076398B1/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/02Knowledge representation; Symbolic representation
    • G06N5/022Knowledge engineering; Knowledge acquisition
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/06Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling

Definitions

  • the invention relates to a computer-based project assessment tool for schematically matching information into a project scheme.
  • the above-described deficiencies are overcome by a system and method adapted for use on a computer platform that provides an ontology that links objects and is capable of being mined.
  • the ontology is comprised of a project ontology framework, a matching engine and a project status matrix that illustrates a multi- relational view of the project status, of confidence levels, or interdiction points and/or positions on project timelines.
  • FIG. 1 illustrates an exemplary project ontology framework according to the present invention
  • FIG. 2 illustrates an exemplary mapping engine according to the present invention
  • FIG. 3 illustrates a block diagram of an Echo State Network
  • FIGS. 4A and 4B respectively illustrate exemplary display status matrices according to the present invention
  • FIG. 5 is a flow chart illustrating the preferred embodiment of the invention.
  • FIG. 6 is an example of a project ontology for the aerospace industry.
  • This invention is adapted for use on a host computer platform including on a personal computer, on a server, on a website, on a local or wide area network, on a PDA or on any other processor-based device known or used in the art.
  • the assessment tool or ontology of the present invention is an explicit formal specification of how to represent objects, concepts and other entities that are assumed to exist in an area of interest.
  • the ontology links the objects and concepts with the relationships among the objects and concepts.
  • the premise is based on the fact that today's organizations, whether government, commercial or otherwise, can be seen as project management specialists that oversee a diverse portfolio of related projects and these projects may share similarities.
  • the ontology project framework uses a template for searching data. Instead of gathering data piece by piece, the template can encompass a whole data set seamlessly.
  • the ontology uses specification tools, such as a Resource Definition Framework (RDF) and an Ontology Web Language (OWL), which allow complete areas of knowledge to be machine processed.
  • RDF Resource Definition Framework
  • OWL Ontology Web Language
  • Neural networks such as Echo State Networks (ESN)
  • ESN Echo State Networks
  • WordNet powerful lexical dictionaries, such as WordNet ® allow more accurate statistical natural language processing (NLP).
  • Any ontology may include a project portfolio.
  • the system is set up to map or mine information to a particular project or activity within the ontology.
  • the ontology matching scheme described herein has been contemplated in a number of different implementation applications.
  • the ontology can be used to predict terrorist or ontology "projects" for project management of important projects or even strategic initiatives in a corporate environment.
  • the ontology can be used for any application where there are relevant projects.
  • the ontology is used to predict terrorist threats.
  • This implementation of the ontology is logical because terrorist organizations behave in a hierarchical structure.
  • the actions or acts that terrorists carry out are projects.
  • terrorist projects can be predicted by analyzing the recent intelligence in conjunction with the terrorist project plan templates or maps.
  • the ontology allows systematic categorization of knowledge concerning a terrorist organization, a learning neural network of terrorist projects, high confidence predictions of project status, and threat warning and intervention.
  • Another example is an ontology that is implemented in a commercial setting.
  • email, instant messaging and even computer keystrokes could be used to feed data into the ontology scheme.
  • the email or other communication messages are used by employees to discuss work activities and the progress of such activities.
  • the casual communication between employees is invaluable to the business.
  • the ontology allows real time project status from mined documents, e-mail, instant messaging, and other project related data repositories, high confidence scenarios for risk management, and greater success in large complex projects.
  • the ontology is comprised of three major components which are described in further detail in the attached figures.
  • the project ontology framework (POF) is the first component.
  • the project ontology includes template projects, discrete project activities, project roles, project template lexicons, project template lexicon networks, sponsor organization lexicons, sponsor organization social networks, sponsor organization lexicon networks, inter sponsor organization social networks, inter sponsor organizational lexicon networks, project template activity networks, inter project template activity networks, sponsor organization project portfolios, etc.
  • the POF is a set of methods for constructing a project ontology.
  • the second component is the matching engine.
  • the matching engine provides a method of associating a given piece of data with a discrete project activity or project template.
  • the third component of the ontology based system is a project status matrix that provides a multi-relational view display of a sponsoring organization by project templates, current assessment matching to show current status, confidence levels, interdiction points, and position on the project timeline.
  • the ontology schematic matching system is constructed using a machine readable language (e.g., OWL).
  • OWL for example, is a specification published by the world wide web consortium (w3c.org). OWL is designed to be used in those environments where the content of information is being processed, not just presented. OWL allows for improved machine interpretability of Web content by providing additional vocabulary along with formal semantics.
  • OWL has three increasingly expressive sublanguages, including OWL Lite, OWL DL and OWL Full.
  • Each POF is constructed of five major classes and their subclasses.
  • the ontology base framework is created from expert input, historical data, what-if exercises, analysis of other ontology states, as well as creative brainstorming.
  • the classes include lexicon, portfolios, project templates, activity and role. Each of these classes has relationship ties to other classes by a class value.
  • FIG. 1 illustrates an exemplary POF.
  • the first class of the POF 182 is the lexicon 12.
  • the lexicon class 12 provides a knowledge base about a subset of words in the vocabulary of a natural language.
  • the lexicon subclasses includes the WordNet ® database, and entity and specialized lexicons mapped to the WordNet ® database.
  • the organization and specialized lexicons are appropriately linked to the classes to which the particularized lexicons apply through lexicon networks.
  • the lexicon class 12 includes lexicon networks which are networks/matrices of words, by use, networked grammatically/cognitively. Lexicon networks are constructed at the portfolio level as well as per project template, activity and role.
  • WordNet ® is an open source application developed by and made available through Princeton University. In WorldNet ® , nouns, verbs, adjectives, adverbs, etc., are organized into sets, each set representing one underlying lexical concept. The word sets are linked by different relations. Although, this invention is contemplated using WordNet ® software, other lexical reference systems or tools inspired by psycholinguistic theories of human lexical memory can be used to develop the Lexicon.
  • the second class of the POF 182 is the portfolio 14.
  • the portfolio 14 includes the subclasses of sponsor(s), projects (templates and ongoing projects), roles and lexicons not clearly associated with a project or an activity, related portfolios and other metadata. Role and activity information not clearly associated with a project or activity is kept in the portfolio structure 182 for later use.
  • the third class of the POF 182 is the project template 16.
  • the project template 16 is made up of several subclasses.
  • the first subclass delineates the sponsor organization.
  • the sponsor organization is the group which is carrying out a particular activity (e.g., Hamas, a corporate competitor, an organization). Activities of the project 16 make up the second subclass.
  • the activities are the different activities that are being carried out or need to be carried out to complete the project.
  • the subclasses also include lexicons and roles that are not clearly associated with an activity. By maintaining this information, if at a later point the associated activity becomes clear, the data may be mapped to the appropriate activity.
  • the project templates allow for information relating to other related projects.
  • the final subclass of the project templates is the other metadata.
  • the fourth class in the POF 182 is activities 18.
  • the activity class 18 is comprised of verbs, nouns, adjectives, adverbs and roles. Additional subclasses include a time sequence of events or actions, related activities and other metadata.
  • the final class in the POF 182 is the role 20.
  • the role class 20 is comprised of the subclasses of skills, functions, command relationship(s) (organizational chart level), tools, named individuals acting in this role, related roles, and other metadata.
  • the base POF 182 is the complete POF structure without any mapping of data. It is possible that the classes of the base POF 182 can change depending on changes to the relationship data and to the measured activity.
  • the base POF 182 is continuously evolving as new relationship data is added and the fidelity in the base POF 182 increases as new relationship data is added. Changes to the base POF 182 are based on the guiding configuration management principals, policies and thresholds. By adhering to guiding configuration management principles, the base POF 182 cannot be changed at the whim of a user.
  • the POF 182 is a large network of relationships codified in machine readable language. Real time instances of the base POF 182 are created by combining a recurrent neural network algorithm (RNN), such as the Echo State Network (ESN) 206 (illustrated in FIG. 3) , with the base POF 182. As data is input into the ontology, the RNN algorithm is constantly generating echoes of the base POF 182. Based on weight (w) and confidence cutoffs, certain echoes are captured.
  • RNN recurrent neural network algorithm
  • ESN Echo State Network
  • non-base POFs There are three basic types of non-base POFs: candidate, working and operational POFs (see FIG. 2).
  • the operational POF is the only POF that is used for status display and base POF refinement.
  • the operational POF also represents the highest confidence echo based on the data that is mapped to the POFs. Information in the operational POF can and often is used to refine the base POF.
  • the difference between candidate, working and operational POF is based on confidence cutoffs.
  • a working POF does not replace the operational POF unless a user decides that it is a more accurate view of the POF.
  • the POF status is based on confidence cutoffs, not every echo will rise to the level of a working POF.
  • echoes that do not rise to the level of working POFs and working POFs can be used later to backtrack and determine if information that has been mapped to a project, role, activity, etc., is still accurate.
  • mapping engine 100 one of the main features of the ontology is a mapping engine 100. Regardless of where the mapping engine 100 maps intelligence data, the actions of the engine 100 remain generally the same. An exception is in the fidelity of the map.
  • the mapping engine 100 carries out six functions, as illustrated in FIG. 2. These six functions include data receipt 102, routing 104, parsing and formatting 106, mapping 108, echo generation (state) 110 and echo maintenance 112.
  • the mapping engine 100 first classifies at 120 incoming intelligence based on the source of the information at 122.
  • the source of the intelligence/data is used to determine which pathway of algorithms the data will travel in preparation for mapping. For example, unstructured data will undergo statistical Natural Language Processing (NLP), while machine-tagged data will go to a transformation function prior to mapping. The data will be processed into the proper RDF/OWL format before mapping.
  • NLP Natural Language Processing
  • the source of the data is used to create message types 122.
  • Exemplary sources of data may include the internet, workflow, email, instant messaging, a document management server, or a project server.
  • Each message type is governed by a rules engine 124 that provides subsequent processing. Processing can include, but is not limited to, evaluation of the source and message formatting (e.g., formatting the message into the proper OWL/RDF format).
  • the rules engine 124 may also provide an initial scoring of the message.
  • the confidence levels assigned throughout the POF ontology will be aggregated to classify the POFs as candidate 160, working 162 and operational 164 POFs. A confidence level is assigned based on the source of the data.
  • the rules engine 124 is in effect, a series of software agents that reside near the repository or source of a message. Aggregate confidence levels are used in echo generation to determine which POFs are candidate, working and operational 160, 162 and 164 respectively.
  • Messages are classified into types, as illustrated in FIG. 2.
  • the message types include "trusted parsed directed” 130, “trusted parsed process” 132, “assigned confidence” 126, “no assigned confidence” 134 and “fragment” 136.
  • the type of intelligence with the highest confidence type is "trusted parsed directed" 130.
  • This message type is from a very high confidence source (e.g., workflow).
  • the data is already parsed using WordNet ® , is in the proper RDF/OWL format, and is directed via routing 142 to a specific area of the ontology.
  • the message is directed to a specific place in the ontology by an end user or is automatically directed because of the source.
  • the message is created by an analyst based on multiple sources to provide a finished mapping of data directly to the highest confidence working POF or all of the POFs.
  • This message type bypasses the parsing 106 and mapping 108 functions of the mapping engine 100.
  • the second message type is "trusted parsed process" 132.
  • This message type is from a high confidence source, usually the end user.
  • the message has already been parsed using WordNet ® , is in the proper RDF/OWL format, and is sent directly through routing module 144 to the mapping engine 108, skipping the parsing function 106.
  • the third message type is "assigned confidence" 126.
  • This message type is assigned a confidence route 140 by source or end user. However, it is a raw piece of data without parsing 106 or formatting (e.g., internet email). The message will proceed through 140 to the parsing function 106, as illustrated in FIG.2, to be processed as described in more detail below.
  • the fourth message type "no assigned confidence" 134 is a raw piece of data without parsing 106 or formatting, but the source is known. Further, whether by source or the end user's direction, the data does not have an assigned confidence level 146. Subsequent processing will be used to determine the confidence level. The message will proceed through routing module 146 to the parsing function 106, as illustrated in FIG. 2.
  • fragment 136 is also raw data without parsing 106 or formatting. However, the source of the data is unknown. Fragment data is weighted lower 148 in the subsequent confidence level processing routing. The message will proceed through routing module 148 therefore to the parsing functionlO ⁇ as illustrated in FIG. 2.
  • the data is routed either to the parsing and formatting 106 function, the mapping function 108, or directly into an echo(es) states 112.
  • FIG. 5 illustrates exemplary steps used for providing lexical processing as described with reference to the data receipt module 102 and routing module 104 in FIG. 2.
  • new data is provided to the mapping engine. That data is then tested at 402 to determine whether or not it matches the POF lexicon. If it matches at 405, then it is next determined whether the data is associated with a project at 406. If there is no matching project at 407, then the data is not entered and/or associated with the POF. [0049] For data that is then successfully associated with a project 409, that data is then tested at 410 to determine whether or not there is an activity that can be associated with that data. An example of a type of activity is a task.
  • Data that is not successfully associated with a project 411, at 408 the data is associated with the POF. Also, data that is not associated successfully associated with an activity at 417 is associated with the identified project at 414.
  • data can be associated with a task at 413
  • the data is then associated with all appropriate tasks and it is determined whether there is metadata to associate with the data. If matching metadata is available at 415, then the new data is also associated at 418 with the metadata. Data that is not matched to metadata at 419, is then associated with the tasks in step 416.
  • the parsing function 106 uses statistical NLP to parse words within a message. Words in a message are tagged 150 as to the part of speech, definition and time 154 (e.g., today, yesterday, September 11, 2001, two weeks, etc.).
  • the part of speech tagging 150 attempts to determine if a word is a noun, verb, determiner, adjective, adverb, pronoun, preposition, particle, conjunction, or number.
  • WordNet ® Lookup, or other similar function in a comparable reference system is performed for each of the words in a message.
  • WordNet ® determines that the word has only one use (e.g., noun)
  • the word is sent to sense processing 152.
  • sense determination 152 may play a deciding factor in determining the word's part of speech.
  • Brill Tagger is an open source application developed by Eric Brill and made available through www.cs.jhu.edu.
  • a modified version of the open source Brill Tagger can be used as one means of part-of-speech tagging 150. Any other conventionally known speech tagging software or hardware, however, can be employed in conjunction with the present invention.
  • Modified Brill Tagger is a modified version of the open source Brill Tagger coupled with WordNet ® and sponsor lexicons. Based on rules associated with the transformation process and the polysemy counts of the associated word, the tags are given a confidence number for later evaluation. A tree traversal, statistical approach of WordNet ® scores each part-of- speech for the words.
  • a tree traversal traverses the chain of concepts (synonyms and related concepts) in the WordNet ® database looking for common works in the message and the change of concepts to determine the sense.
  • the tree traversal processing also determines the sense of the word 152. Both the tag 150 and the sense 152 are scored with a confidence 156 for final processing.
  • Wording mapping including tagging 150 and sense determination 152 works as follows.
  • the message is transformed into a series of networks by connecting words that are used in the same sentence, phrase, and clause (i.e., creating a words in context network).
  • a master network is created by all of the sentence, phrase, and clause networks. Multiple uses of the same word in the message phrasing are depicted in the strength of the connection.
  • the networks are actually stored as matrices of all of the words, as both row and column. The strength of connection between words is signified as a number.
  • the algorithm next looks to the existing sponsor, project, and activity lexicons, or other training corpuses, to seek a similar sentence, phrase, and clause matrix. The entire message matrix is used to seek matching matrices in the project and activity lexicon matrices.
  • the highest confidence match is used to tag the part of speech 150, and select the appropriate sense 152 for each word in the message. A confidence score is applied for later processing. [0057] The confidence scores are used to de-conflict differences in tag and sense selections using a combination of voting and highest confidence wins process.
  • each word is time context processed 154.
  • Time context processing is a specialized form of sense determination.
  • Each word is evaluated to determine if it's sense is related to time.
  • the sense provided by the parsing function 106 is used to traverse through the chain of concepts in the WordNet ® database or databases created by software having similar features to WordNet. If time related words/concepts are detected, an algorithm attempts to place the word on a continuous time-line. If possible, the word is given a distinct date, otherwise the word is given the appropriate level calendar fidelity.
  • the message is evaluated to extract and format the data according to the main areas of the ontology for mapping 108.
  • the message is formatted according to the following roles at 180:
  • Portfolio Format when more than one portfolio is present.
  • Lexicon Format to determine sponsor organization, project template, or activity.
  • Project Template Format to determine which project template.
  • Activity Format to determine the appropriate activity(ies) present in the message.
  • Role Format to determine the appropriate role(s) present in the message.
  • Each format allows the mapping process 108 or module 180 to step through the message from multiple perspectives.
  • the mapping process 108 or module 180 generates an echo or state 110 for each format, as well as one encompassing all of the formats.
  • the reformatted message is input to at least one RNN algorithm, as illustrated in FIG. 3 (element 200) during the mapping module or process.
  • Each project, activity, role, etc. may each have a RNN algorithm to process the inputted message.
  • the RNN algorithm tests the input message against the base POF 182 (FIG. 2).
  • Each item within the POF has a threshold value for an output 184 to modify a candidate POF 160, working POF 162 and operational POF 164 status.
  • a message may have only one high confidence match within the base POF 182, or it may have multiple matches (it may match multiple activities for example).
  • a high confidence match is user definable, but would generally be considered as statistically significant.
  • an activity format message is fed into the module or process 180, it is evaluated against all of the activity internals based on match weight (see FIG. 3, element W).
  • each activity word is evaluated to the activity lexicon including synonyms, antonyms, meronyms, holonyms, hypernyms, and sense.
  • An aggregate score for each word is assigned based on the probability of the activity word matching the activity lexicon in the base POF 182.
  • the activity lexicon network/matrix of the message is evaluated against the activity lexicon network/matrix in the base POF 182.
  • the activity lexicon matrix is a networking of words that are linked from a grammatical and cognitive relationship. This evaluation is also given a score. The score represents the statistical probability that the message activity matrix matches the activity matrix in the POF. From there, each activity attribute is evaluated against each activity in the POF (e.g., time sequence, role, etc.) 190. If the overall score exceeds the match weight, then the match moves to the second algorithm. As the base POF 182 is refined, the match weights increase, thus increasing the confidence level of each POF state 112.
  • the project portfolio status matrix provides a visual display of the operational POF. Visualization can be accomplished on any known type of visualization hardware including but not limited to a CRT, flat panel projection, LCD or LED matrix, etc.
  • FIGS. 4A and 4B illustrate exemplary status matrices according to the present invention.
  • the status is displayed by the sponsoring organization portfolio.
  • the matrix provides a visualization of the projects based on the POF confidence level, and possible interdiction opportunities.
  • the timelines maybe depict the status of the projects over a set period of time or may be adjustable for a specific time view.
  • FIG. 4A illustrates an exemplary status matrix 300 of the operational POFs that can exist in a commercial application.
  • FIG. 4B illustrates an exemplary status matrix 350 of the operation POF that can exist if the project is based on terrorist projects.
  • the matrix can provide information on projects that are jointly sponsored.
  • the time lines may be depicted by a variety of colors to illustrate which projects are nearing completion or interdictions point, for example.
  • the vertical axis illustrates several typical commercial applications or types of projects 320 illustrated by the matrix 300.
  • the horizontal matrix in this example shows different typical company offices or performing organizations 330 that relate to each gant bar 336.
  • the offices represented include the company CIO (chief information officer), company CMO (chief marketing officer), company CFO (chief financial officer), and company CTO (chief technical officer).
  • the gant bar 336 is subdivided by time-based markers 332.
  • the top marker 334 illustrates current project status based on the time line 332.
  • the internal markers 338 represent a high risk area for a company. At this point, this is typically deemed a point of weakness where projects are most likely to fail or be delayed or suffer cost overruns.
  • the same gant bar display 330 is applied to examples of the different types of projects (e.g. vehicular bomb, biological and radiological) for different types of performing organizations 330 (e.g. Hamas, Al gori, Islamic Jihad and Domestic).
  • the interdiction point, or points 338 now represent areas where police or military intervention is optimal, and where these projects have their greatest weakness and exposure points.
  • FIG. 6 illustrates an exemplary project ontology, according to the afore-described invention, providing the intersection of two taxonomies.

Abstract

A computerized project management analytical system and method that develops and manages an ontology that links objects and is capable of being mined. The ontology is comprised of a project ontology framework, a matching engine and a project status matrix that illustrates a multi-relational view of the project status, of confidence levels, or interdiction points and/or positions on project timelines.

Description

PREDICTIVE ANALYTIC METHOD AND APPARATUS
[0001] This application claims benefit to U.S. Provisional Application 60/642, 983, which is herein incorporated by reference.
FIELD OF THE INVENTION
[0002] The invention relates to a computer-based project assessment tool for schematically matching information into a project scheme.
BACKGROUND OF THE INVENTION
[0003] Today, in project management, the focus of analysis and control is on the ability to estimate and associate what is effectively remembered as important with a given project. In other words, since seventy percent of all projects fail based on their original budget or finish date, it is clear that current systems struggle with successful estimations for outcomes. Part of this failure to predict, analyze and control project outcome stems from the inability to effectively mine and place into the proper context the avalanche of the data that could positively improve the predictive outcome of the project.
[0004] Project management, search software, data mining software and statistical/analytical tools could be used resolve project management shortfalls. However, these various tools exist in their own silos and are thereby not associated in a meaningful and usable manner. This failure is exacerbated as the complexity of projects increases as technology and society evolve.
[0005] Moreover, the concept of a project for many human endeavors is becoming widespread and mutating so that increasingly sophisticated tools, if applied correctly, could be implemented in more wide-ranging environments. For example, tools could be used in different ways, depending on the wide range of possibilities of what constitutes a "a project", and who is the "project manager". For example, a terrorist planning his or her attack could be a deemed a "project manager" in the same way a more traditional individual, such as a certified project engineer, could plan a construction, research or information technology project. Other environments that rely heavily on project management and control and that could benefit from a more sophisticated analytical approach to project management include but are not limited to the film industry, the automotive industry, advertising, drug/pharmaceutical research, clinical medical trials, to name a few.
[0006] A need therefore exists in the art for a predictive analytic system and method that employs the best available software tools and that run on standard computer hardware in order to provide project predictive analytics to the end user.
SUMMARY OF THE INVENTION
[0007] The above-described deficiencies are overcome by a system and method adapted for use on a computer platform that provides an ontology that links objects and is capable of being mined. The ontology is comprised of a project ontology framework, a matching engine and a project status matrix that illustrates a multi- relational view of the project status, of confidence levels, or interdiction points and/or positions on project timelines.
BRIEF DESCRIPTION OF THE DRAWINGS
[0008] FIG. 1 illustrates an exemplary project ontology framework according to the present invention;
[0009] FIG. 2 illustrates an exemplary mapping engine according to the present invention;
[0010] FIG. 3 illustrates a block diagram of an Echo State Network; [0011] FIGS. 4A and 4B respectively illustrate exemplary display status matrices according to the present invention;
[0012] FIG. 5 is a flow chart illustrating the preferred embodiment of the invention; and
[0013] FIG. 6 is an example of a project ontology for the aerospace industry.
DETAILED DESCRIPTION OF THE INVENTION
[0014] The present invention described in the following specification and in the attached drawings wherein like elements are referenced to like reference numerals.
[0015] This invention is adapted for use on a host computer platform including on a personal computer, on a server, on a website, on a local or wide area network, on a PDA or on any other processor-based device known or used in the art.
[0016] The assessment tool or ontology of the present invention is an explicit formal specification of how to represent objects, concepts and other entities that are assumed to exist in an area of interest. The ontology links the objects and concepts with the relationships among the objects and concepts. The premise is based on the fact that today's organizations, whether government, commercial or otherwise, can be seen as project management specialists that oversee a diverse portfolio of related projects and these projects may share similarities.
[0017] Not only does the ontology mine data but it provides current status, while making predictions about future status. In its broadest sense, the ontology project framework uses a template for searching data. Instead of gathering data piece by piece, the template can encompass a whole data set seamlessly.
[0018] Specification tools, neural network technology and natural language processing are needed to create an effective ontology. The ontology uses specification tools, such as a Resource Definition Framework (RDF) and an Ontology Web Language (OWL), which allow complete areas of knowledge to be machine processed. Neural networks, such as Echo State Networks (ESN), simplify machine processing while increasing accuracy. In addition, powerful lexical dictionaries, such as WordNet® allow more accurate statistical natural language processing (NLP).
[0019] Objects and concepts are mined in projects. Any ontology may include a project portfolio. As will be described in the present specification, the system is set up to map or mine information to a particular project or activity within the ontology. The ontology matching scheme described herein has been contemplated in a number of different implementation applications. For example, the ontology can be used to predict terrorist or ontology "projects" for project management of important projects or even strategic initiatives in a corporate environment.
[0020] The ontology can be used for any application where there are relevant projects. For example, the ontology is used to predict terrorist threats. This implementation of the ontology is logical because terrorist organizations behave in a hierarchical structure. In the ontology, the actions or acts that terrorists carry out are projects. Using this structure and additional intelligence mapped into the ontology structure, terrorist projects can be predicted by analyzing the recent intelligence in conjunction with the terrorist project plan templates or maps. In an intelligence analysis perspective, the ontology allows systematic categorization of knowledge concerning a terrorist organization, a learning neural network of terrorist projects, high confidence predictions of project status, and threat warning and intervention.
[0021] Another example is an ontology that is implemented in a commercial setting. In one example, email, instant messaging and even computer keystrokes could be used to feed data into the ontology scheme. The email or other communication messages are used by employees to discuss work activities and the progress of such activities. The casual communication between employees is invaluable to the business. By tracking and using the communications, the progress of the strategic initiative and important projects can be predicted. The prediction is made by analyzing the corporate communication, against strategic alignment or project template plans. In the commercial setting, the ontology allows real time project status from mined documents, e-mail, instant messaging, and other project related data repositories, high confidence scenarios for risk management, and greater success in large complex projects. These implementation examples may be used throughout the specification to provide a context for the ontology scheme.
[0022] The ontology is comprised of three major components which are described in further detail in the attached figures. The project ontology framework (POF) is the first component. The project ontology includes template projects, discrete project activities, project roles, project template lexicons, project template lexicon networks, sponsor organization lexicons, sponsor organization social networks, sponsor organization lexicon networks, inter sponsor organization social networks, inter sponsor organizational lexicon networks, project template activity networks, inter project template activity networks, sponsor organization project portfolios, etc. The POF is a set of methods for constructing a project ontology.
[0023] The second component is the matching engine. The matching engine provides a method of associating a given piece of data with a discrete project activity or project template.
[0024] The third component of the ontology based system is a project status matrix that provides a multi-relational view display of a sponsoring organization by project templates, current assessment matching to show current status, confidence levels, interdiction points, and position on the project timeline. [0025] The ontology schematic matching system is constructed using a machine readable language (e.g., OWL). OWL, for example, is a specification published by the world wide web consortium (w3c.org). OWL is designed to be used in those environments where the content of information is being processed, not just presented. OWL allows for improved machine interpretability of Web content by providing additional vocabulary along with formal semantics. OWL has three increasingly expressive sublanguages, including OWL Lite, OWL DL and OWL Full.
Project Ontology Framework (POF)
[0026] Each POF is constructed of five major classes and their subclasses. The ontology base framework is created from expert input, historical data, what-if exercises, analysis of other ontology states, as well as creative brainstorming. The classes include lexicon, portfolios, project templates, activity and role. Each of these classes has relationship ties to other classes by a class value. FIG. 1 illustrates an exemplary POF.
[0027] The first class of the POF 182 is the lexicon 12. The lexicon class 12 provides a knowledge base about a subset of words in the vocabulary of a natural language. The lexicon subclasses includes the WordNet® database, and entity and specialized lexicons mapped to the WordNet® database. The organization and specialized lexicons are appropriately linked to the classes to which the particularized lexicons apply through lexicon networks. Additionally, the lexicon class 12 includes lexicon networks which are networks/matrices of words, by use, networked grammatically/cognitively. Lexicon networks are constructed at the portfolio level as well as per project template, activity and role.
[0028] WordNet® is an open source application developed by and made available through Princeton University. In WorldNet®, nouns, verbs, adjectives, adverbs, etc., are organized into sets, each set representing one underlying lexical concept. The word sets are linked by different relations. Although, this invention is contemplated using WordNet® software, other lexical reference systems or tools inspired by psycholinguistic theories of human lexical memory can be used to develop the Lexicon.
[0029] The second class of the POF 182 is the portfolio 14. The portfolio 14 includes the subclasses of sponsor(s), projects (templates and ongoing projects), roles and lexicons not clearly associated with a project or an activity, related portfolios and other metadata. Role and activity information not clearly associated with a project or activity is kept in the portfolio structure 182 for later use.
[0030] The third class of the POF 182 is the project template 16. The project template 16 is made up of several subclasses. The first subclass delineates the sponsor organization. The sponsor organization is the group which is carrying out a particular activity (e.g., Hamas, a corporate competitor, an organization). Activities of the project 16 make up the second subclass. The activities are the different activities that are being carried out or need to be carried out to complete the project. The subclasses also include lexicons and roles that are not clearly associated with an activity. By maintaining this information, if at a later point the associated activity becomes clear, the data may be mapped to the appropriate activity. Additionally, the project templates allow for information relating to other related projects. The final subclass of the project templates is the other metadata.
[0031] The fourth class in the POF 182 is activities 18. The activity class 18 is comprised of verbs, nouns, adjectives, adverbs and roles. Additional subclasses include a time sequence of events or actions, related activities and other metadata.
[0032] The final class in the POF 182 is the role 20. The role class 20 is comprised of the subclasses of skills, functions, command relationship(s) (organizational chart level), tools, named individuals acting in this role, related roles, and other metadata.
[0033] The base POF 182 is the complete POF structure without any mapping of data. It is possible that the classes of the base POF 182 can change depending on changes to the relationship data and to the measured activity. The base POF 182 is continuously evolving as new relationship data is added and the fidelity in the base POF 182 increases as new relationship data is added. Changes to the base POF 182 are based on the guiding configuration management principals, policies and thresholds. By adhering to guiding configuration management principles, the base POF 182 cannot be changed at the whim of a user.
[0034] The POF 182 is a large network of relationships codified in machine readable language. Real time instances of the base POF 182 are created by combining a recurrent neural network algorithm (RNN), such as the Echo State Network (ESN) 206 (illustrated in FIG. 3) , with the base POF 182. As data is input into the ontology, the RNN algorithm is constantly generating echoes of the base POF 182. Based on weight (w) and confidence cutoffs, certain echoes are captured.
[0035] There are three basic types of non-base POFs: candidate, working and operational POFs (see FIG. 2). The operational POF is the only POF that is used for status display and base POF refinement. The operational POF also represents the highest confidence echo based on the data that is mapped to the POFs. Information in the operational POF can and often is used to refine the base POF. The difference between candidate, working and operational POF is based on confidence cutoffs. A working POF does not replace the operational POF unless a user decides that it is a more accurate view of the POF. Additionally, because the POF status is based on confidence cutoffs, not every echo will rise to the level of a working POF. [0036] However, echoes that do not rise to the level of working POFs and working POFs can be used later to backtrack and determine if information that has been mapped to a project, role, activity, etc., is still accurate.
Mapping Engine/Data Receipt & Routing
[0037] Referring now to FIG. 2, one of the main features of the ontology is a mapping engine 100. Regardless of where the mapping engine 100 maps intelligence data, the actions of the engine 100 remain generally the same. An exception is in the fidelity of the map. The mapping engine 100 carries out six functions, as illustrated in FIG. 2. These six functions include data receipt 102, routing 104, parsing and formatting 106, mapping 108, echo generation (state) 110 and echo maintenance 112.
[0038] The mapping engine 100 first classifies at 120 incoming intelligence based on the source of the information at 122. The source of the intelligence/data is used to determine which pathway of algorithms the data will travel in preparation for mapping. For example, unstructured data will undergo statistical Natural Language Processing (NLP), while machine-tagged data will go to a transformation function prior to mapping. The data will be processed into the proper RDF/OWL format before mapping.
[0039] The source of the data is used to create message types 122. Exemplary sources of data may include the internet, workflow, email, instant messaging, a document management server, or a project server. Each message type is governed by a rules engine 124 that provides subsequent processing. Processing can include, but is not limited to, evaluation of the source and message formatting (e.g., formatting the message into the proper OWL/RDF format). The rules engine 124 may also provide an initial scoring of the message. The confidence levels assigned throughout the POF ontology will be aggregated to classify the POFs as candidate 160, working 162 and operational 164 POFs. A confidence level is assigned based on the source of the data. The rules engine 124 is in effect, a series of software agents that reside near the repository or source of a message. Aggregate confidence levels are used in echo generation to determine which POFs are candidate, working and operational 160, 162 and 164 respectively.
[0040] Messages are classified into types, as illustrated in FIG. 2. The message types include "trusted parsed directed" 130, "trusted parsed process" 132, "assigned confidence" 126, "no assigned confidence" 134 and "fragment" 136.
[0041] The type of intelligence with the highest confidence type is "trusted parsed directed" 130. This message type is from a very high confidence source (e.g., workflow). The data is already parsed using WordNet®, is in the proper RDF/OWL format, and is directed via routing 142 to a specific area of the ontology. The message is directed to a specific place in the ontology by an end user or is automatically directed because of the source. Optionally, the message is created by an analyst based on multiple sources to provide a finished mapping of data directly to the highest confidence working POF or all of the POFs. This message type bypasses the parsing 106 and mapping 108 functions of the mapping engine 100.
[0042] The second message type is "trusted parsed process" 132. This message type is from a high confidence source, usually the end user. The message has already been parsed using WordNet®, is in the proper RDF/OWL format, and is sent directly through routing module 144 to the mapping engine 108, skipping the parsing function 106.
[0043] The third message type is "assigned confidence" 126. This message type is assigned a confidence route 140 by source or end user. However, it is a raw piece of data without parsing 106 or formatting (e.g., internet email). The message will proceed through 140 to the parsing function 106, as illustrated in FIG.2, to be processed as described in more detail below.
[0044] The fourth message type "no assigned confidence" 134 is a raw piece of data without parsing 106 or formatting, but the source is known. Further, whether by source or the end user's direction, the data does not have an assigned confidence level 146. Subsequent processing will be used to determine the confidence level. The message will proceed through routing module 146 to the parsing function 106, as illustrated in FIG. 2.
[0045] The last message type, fragment 136, is also raw data without parsing 106 or formatting. However, the source of the data is unknown. Fragment data is weighted lower 148 in the subsequent confidence level processing routing. The message will proceed through routing module 148 therefore to the parsing functionlOό as illustrated in FIG. 2.
[0046] Based on the message type and configurable rules 124 of the ontology framework, the data is routed either to the parsing and formatting 106 function, the mapping function 108, or directly into an echo(es) states 112.
[0047] FIG. 5 illustrates exemplary steps used for providing lexical processing as described with reference to the data receipt module 102 and routing module 104 in FIG. 2.
[0048] At 400, new data is provided to the mapping engine. That data is then tested at 402 to determine whether or not it matches the POF lexicon. If it matches at 405, then it is next determined whether the data is associated with a project at 406. If there is no matching project at 407, then the data is not entered and/or associated with the POF. [0049] For data that is then successfully associated with a project 409, that data is then tested at 410 to determine whether or not there is an activity that can be associated with that data. An example of a type of activity is a task.
[0050] Data that is not successfully associated with a project 411, at 408 the data is associated with the POF. Also, data that is not associated successfully associated with an activity at 417 is associated with the identified project at 414.
[0051] Referring back to 410, if data can be associated with a task at 413, the data is then associated with all appropriate tasks and it is determined whether there is metadata to associate with the data. If matching metadata is available at 415, then the new data is also associated at 418 with the metadata. Data that is not matched to metadata at 419, is then associated with the tasks in step 416.
Mapping Engine - Parsing
[0052] The parsing function 106 uses statistical NLP to parse words within a message. Words in a message are tagged 150 as to the part of speech, definition and time 154 (e.g., today, yesterday, September 11, 2001, two weeks, etc.).
[0053] The part of speech tagging 150 attempts to determine if a word is a noun, verb, determiner, adjective, adverb, pronoun, preposition, particle, conjunction, or number. In determining the part of speech, WordNet® Lookup, or other similar function in a comparable reference system, is performed for each of the words in a message. When WordNet® determines that the word has only one use ( e.g., noun), the word is sent to sense processing 152. However, if a word has multiple parts of speech, sense determination 152 may play a deciding factor in determining the word's part of speech.
[0054] Brill Tagger is an open source application developed by Eric Brill and made available through www.cs.jhu.edu. A modified version of the open source Brill Tagger can be used as one means of part-of-speech tagging 150. Any other conventionally known speech tagging software or hardware, however, can be employed in conjunction with the present invention. Modified Brill Tagger is a modified version of the open source Brill Tagger coupled with WordNet® and sponsor lexicons. Based on rules associated with the transformation process and the polysemy counts of the associated word, the tags are given a confidence number for later evaluation. A tree traversal, statistical approach of WordNet® scores each part-of- speech for the words. A tree traversal traverses the chain of concepts (synonyms and related concepts) in the WordNet® database looking for common works in the message and the change of concepts to determine the sense. The tree traversal processing also determines the sense of the word 152. Both the tag 150 and the sense 152 are scored with a confidence 156 for final processing.
[0055] Wording mapping, including tagging 150 and sense determination 152 works as follows. The message is transformed into a series of networks by connecting words that are used in the same sentence, phrase, and clause (i.e., creating a words in context network). A master network is created by all of the sentence, phrase, and clause networks. Multiple uses of the same word in the message phrasing are depicted in the strength of the connection. The networks are actually stored as matrices of all of the words, as both row and column. The strength of connection between words is signified as a number. The algorithm next looks to the existing sponsor, project, and activity lexicons, or other training corpuses, to seek a similar sentence, phrase, and clause matrix. The entire message matrix is used to seek matching matrices in the project and activity lexicon matrices.
[0056] The highest confidence match is used to tag the part of speech 150, and select the appropriate sense 152 for each word in the message. A confidence score is applied for later processing. [0057] The confidence scores are used to de-conflict differences in tag and sense selections using a combination of voting and highest confidence wins process.
[0058] After the part of speech tagging 150 and sense determination 152 is finished, each word is time context processed 154. Time context processing is a specialized form of sense determination. Each word is evaluated to determine if it's sense is related to time. The sense provided by the parsing function 106 is used to traverse through the chain of concepts in the WordNet® database or databases created by software having similar features to WordNet. If time related words/concepts are detected, an algorithm attempts to place the word on a continuous time-line. If possible, the word is given a distinct date, otherwise the word is given the appropriate level calendar fidelity.
Mapping and State/Echo Generation
[0059] Once part of speech tagging and sense selection is done for each word, the message is evaluated to extract and format the data according to the main areas of the ontology for mapping 108. The message is formatted according to the following roles at 180:
[0060] Portfolio Format - when more than one portfolio is present. Lexicon Format - to determine sponsor organization, project template, or activity. Project Template Format - to determine which project template. Activity Format - to determine the appropriate activity(ies) present in the message. Role Format - to determine the appropriate role(s) present in the message.
[0061] Each format allows the mapping process 108 or module 180 to step through the message from multiple perspectives. The mapping process 108 or module 180 generates an echo or state 110 for each format, as well as one encompassing all of the formats. [0062] The reformatted message is input to at least one RNN algorithm, as illustrated in FIG. 3 (element 200) during the mapping module or process. Each project, activity, role, etc. may each have a RNN algorithm to process the inputted message. Ideally, the RNN algorithm tests the input message against the base POF 182 (FIG. 2). Each item within the POF has a threshold value for an output 184 to modify a candidate POF 160, working POF 162 and operational POF 164 status. A message may have only one high confidence match within the base POF 182, or it may have multiple matches (it may match multiple activities for example). A high confidence match is user definable, but would generally be considered as statistically significant. Once at least one match has been made, at least a second module of process generates an echo or state of the match 190 for each candidate 160, working 162, and the operational POF 164. New overall scores are then calculated, and if the match exceeds the candidate 160, or working 110 thresholds, or outscores the current operational POF, then the echoed state is promoted accordingly by module or process 190 through output 184.
[0063] For example, if an activity format message is fed into the module or process 180, it is evaluated against all of the activity internals based on match weight (see FIG. 3, element W). First, each activity word is evaluated to the activity lexicon including synonyms, antonyms, meronyms, holonyms, hypernyms, and sense. An aggregate score for each word is assigned based on the probability of the activity word matching the activity lexicon in the base POF 182.
[0064] Next, the activity lexicon network/matrix of the message is evaluated against the activity lexicon network/matrix in the base POF 182. The activity lexicon matrix is a networking of words that are linked from a grammatical and cognitive relationship. This evaluation is also given a score. The score represents the statistical probability that the message activity matrix matches the activity matrix in the POF. From there, each activity attribute is evaluated against each activity in the POF (e.g., time sequence, role, etc.) 190. If the overall score exceeds the match weight, then the match moves to the second algorithm. As the base POF 182 is refined, the match weights increase, thus increasing the confidence level of each POF state 112.
Status Matrix
[0065] Referring now to FIGS. 4A & 4B, the project portfolio status matrix provides a visual display of the operational POF. Visualization can be accomplished on any known type of visualization hardware including but not limited to a CRT, flat panel projection, LCD or LED matrix, etc.
[0066] FIGS. 4A and 4B illustrate exemplary status matrices according to the present invention. The status is displayed by the sponsoring organization portfolio. The matrix provides a visualization of the projects based on the POF confidence level, and possible interdiction opportunities. The timelines maybe depict the status of the projects over a set period of time or may be adjustable for a specific time view.
[0067] FIG. 4A illustrates an exemplary status matrix 300 of the operational POFs that can exist in a commercial application. FIG. 4B illustrates an exemplary status matrix 350 of the operation POF that can exist if the project is based on terrorist projects. The matrix can provide information on projects that are jointly sponsored. The time lines may be depicted by a variety of colors to illustrate which projects are nearing completion or interdictions point, for example.
[0068] Referring to FIG. 4 A, the vertical axis illustrates several typical commercial applications or types of projects 320 illustrated by the matrix 300. The horizontal matrix in this example shows different typical company offices or performing organizations 330 that relate to each gant bar 336. In this example, the offices represented include the company CIO (chief information officer), company CMO (chief marketing officer), company CFO (chief financial officer), and company CTO (chief technical officer).
[0069] The gant bar 336 is subdivided by time-based markers 332. The top marker 334 illustrates current project status based on the time line 332. The internal markers 338 represent a high risk area for a company. At this point, this is typically deemed a point of weakness where projects are most likely to fail or be delayed or suffer cost overruns.
[0070] In FIG. 4B, the same gant bar display 330 is applied to examples of the different types of projects (e.g. vehicular bomb, biological and radiological) for different types of performing organizations 330 (e.g. Hamas, Al Qaeda, Islamic Jihad and Domestic). However, the interdiction point, or points 338, now represent areas where police or military intervention is optimal, and where these projects have their greatest weakness and exposure points.
[0071] FIG. 6 illustrates an exemplary project ontology, according to the afore-described invention, providing the intersection of two taxonomies.
[0072] The above description and accompanying figures are only illustrative of exemplary embodiments that can achieve the features and advantages of the present invention. It is not intended that the invention be limited to the embodiments shown and described in detail herein.

Claims

• IN THE CLAIMSWhat is claimed as new and desired to be protected by Letters Patent of the United States is:
1. A method of receiving and classifying data including:
receiving an intelligence message;
classifying said message as to a type of source; and
further classifying said source information to create a message type.
2. A method of receiving and classifying data according to claim 1, wherein classifying said message type includes, identifying the message types as trusted, parsed, directed, trusted parsed processed, assigned confidence, no assigned confidence and fragment.
PCT/US2006/000908 2005-01-12 2006-01-12 Predictive analytic method and apparatus WO2006076398A2 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US64298305P 2005-01-12 2005-01-12
US60/642,983 2005-01-12

Publications (3)

Publication Number Publication Date
WO2006076398A2 true WO2006076398A2 (en) 2006-07-20
WO2006076398A3 WO2006076398A3 (en) 2007-11-22
WO2006076398B1 WO2006076398B1 (en) 2008-01-10

Family

ID=36678148

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2006/000908 WO2006076398A2 (en) 2005-01-12 2006-01-12 Predictive analytic method and apparatus

Country Status (2)

Country Link
US (3) US7822747B2 (en)
WO (1) WO2006076398A2 (en)

Families Citing this family (47)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2007084790A2 (en) 2006-01-20 2007-07-26 Glenbrook Associates, Inc. System and method for context-rich database optimized for processing of concepts
JP4957075B2 (en) * 2006-05-15 2012-06-20 富士通株式会社 Reliability evaluation program and reliability evaluation apparatus
US7769684B2 (en) * 2006-05-19 2010-08-03 Accenture Global Services Gmbh Semi-quantitative risk analysis
US9202184B2 (en) 2006-09-07 2015-12-01 International Business Machines Corporation Optimizing the selection, verification, and deployment of expert resources in a time of chaos
US7752154B2 (en) * 2007-02-26 2010-07-06 International Business Machines Corporation System and method for deriving a hierarchical event based database optimized for analysis of criminal and security information
US20080294459A1 (en) * 2006-10-03 2008-11-27 International Business Machines Corporation Health Care Derivatives as a Result of Real Time Patient Analytics
US8055603B2 (en) 2006-10-03 2011-11-08 International Business Machines Corporation Automatic generation of new rules for processing synthetic events using computer-based learning processes
US8145582B2 (en) * 2006-10-03 2012-03-27 International Business Machines Corporation Synthetic events for real time patient analysis
US20080109406A1 (en) * 2006-11-06 2008-05-08 Santhana Krishnasamy Instant message tagging
US7853611B2 (en) 2007-02-26 2010-12-14 International Business Machines Corporation System and method for deriving a hierarchical event based database having action triggers based on inferred probabilities
US7970759B2 (en) 2007-02-26 2011-06-28 International Business Machines Corporation System and method for deriving a hierarchical event based database optimized for pharmaceutical analysis
US7788203B2 (en) * 2007-02-26 2010-08-31 International Business Machines Corporation System and method of accident investigation for complex situations involving numerous known and unknown factors along with their probabilistic weightings
US7792774B2 (en) 2007-02-26 2010-09-07 International Business Machines Corporation System and method for deriving a hierarchical event based database optimized for analysis of chaotic events
US8332209B2 (en) * 2007-04-24 2012-12-11 Zinovy D. Grinblat Method and system for text compression and decompression
CN101868811B (en) * 2007-09-19 2013-03-06 联合工艺公司 System and method for threat propagation estimation
US7930262B2 (en) * 2007-10-18 2011-04-19 International Business Machines Corporation System and method for the longitudinal analysis of education outcomes using cohort life cycles, cluster analytics-based cohort analysis, and probabilistic data schemas
US7779051B2 (en) 2008-01-02 2010-08-17 International Business Machines Corporation System and method for optimizing federated and ETL'd databases with considerations of specialized data structures within an environment having multidimensional constraints
CA2650348A1 (en) * 2008-01-22 2009-07-22 Accenture Global Services Gmbh Knowledge transfer in a project environment
US8417694B2 (en) * 2008-03-31 2013-04-09 International Business Machines Corporation System and method for constructing targeted ranking from multiple information sources
CN101605141A (en) * 2008-08-05 2009-12-16 天津大学 Web service relational network system based on semanteme
US20100235314A1 (en) * 2009-02-12 2010-09-16 Decisive Analytics Corporation Method and apparatus for analyzing and interrelating video data
US8458105B2 (en) * 2009-02-12 2013-06-04 Decisive Analytics Corporation Method and apparatus for analyzing and interrelating data
US8090770B2 (en) * 2009-04-14 2012-01-03 Fusz Digital Ltd. Systems and methods for identifying non-terrorists using social networking
US10318877B2 (en) 2010-10-19 2019-06-11 International Business Machines Corporation Cohort-based prediction of a future event
US20120101860A1 (en) * 2010-10-25 2012-04-26 Ezzat Ahmed K Providing business intelligence
US20120284200A1 (en) * 2011-05-06 2012-11-08 Brad Pedersen System for computerized management of patent-related information
US9317825B2 (en) 2011-06-27 2016-04-19 Deltek, Inc. System and method for managing projects
US20130060588A1 (en) * 2011-09-06 2013-03-07 International Business Machines Corporation Modeling and monitoring a relationship with a client and assessing the quality of the relationship
US8666982B2 (en) 2011-10-06 2014-03-04 GM Global Technology Operations LLC Method and system to augment vehicle domain ontologies for vehicle diagnosis
EP2836982B1 (en) 2012-03-05 2020-02-05 R. R. Donnelley & Sons Company Digital content delivery
IL219362A (en) * 2012-04-23 2017-04-30 Verint Systems Ltd System and method for prediction of threatened points of interest
US9208440B2 (en) * 2012-05-29 2015-12-08 Battelle Memorial Institute Method of analyzing a scenario represented as elements of a tensor space, and scored using tensor operators
US9317829B2 (en) * 2012-11-08 2016-04-19 International Business Machines Corporation Diagnosing incidents for information technology service management
WO2014203039A1 (en) 2013-06-19 2014-12-24 Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi System and method for implementing reservoir computing using cellular automata
WO2014203038A1 (en) 2013-06-19 2014-12-24 Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi System and method for implementing reservoir computing in magnetic resonance imaging device using elastography techniques
US10867597B2 (en) * 2013-09-02 2020-12-15 Microsoft Technology Licensing, Llc Assignment of semantic labels to a sequence of words using neural network architectures
US9679247B2 (en) 2013-09-19 2017-06-13 International Business Machines Corporation Graph matching
US9922092B2 (en) 2014-04-24 2018-03-20 Canon Kabushiki Kaisha Devices, systems, and methods for context management
US9390142B2 (en) * 2014-06-05 2016-07-12 Sap Se Guided predictive analysis with the use of templates
US10127901B2 (en) 2014-06-13 2018-11-13 Microsoft Technology Licensing, Llc Hyper-structure recurrent neural networks for text-to-speech
CN104363104B (en) * 2014-09-29 2018-02-09 中国人民解放军总参谋部第五十四研究所 A kind of magnanimity multivariate data battle state display System and method for of Users ' Need-oriented
US10515344B1 (en) 2015-02-10 2019-12-24 Open Invention Network Llc Location awareness assistant that activates a business-oriented operation system or a personal-oriented operation system based on conditions
US10636057B2 (en) * 2015-11-13 2020-04-28 [24]7.ai, Inc. Method and apparatus for dynamically selecting content for online visitors
US20190012629A1 (en) * 2017-07-10 2019-01-10 Findo, Inc. Team performance supervisor
US10992763B2 (en) 2018-08-21 2021-04-27 Bank Of America Corporation Dynamic interaction optimization and cross channel profile determination through online machine learning
US11115440B2 (en) 2019-05-08 2021-09-07 Bank Of America Corporation Dynamic threat intelligence detection and control system
US11507908B2 (en) * 2021-03-17 2022-11-22 Accenture Global Solutions Limited System and method for dynamic performance optimization

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030023573A1 (en) * 2001-07-27 2003-01-30 International Business Machines Corporation Conflict-handling assimilator service for exchange of rules with merging
US6529934B1 (en) * 1998-05-06 2003-03-04 Kabushiki Kaisha Toshiba Information processing system and method for same

Family Cites Families (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5111400A (en) * 1987-03-16 1992-05-05 Yoder Evan W Automatic integrated real-time flight crew information system
AU692369B2 (en) * 1995-02-02 1998-06-04 Aprisma Management Technologies, Inc. Method and apparatus for learning network behavior trends and predicting future behavior of communications networks
US6631346B1 (en) * 1999-04-07 2003-10-07 Matsushita Electric Industrial Co., Ltd. Method and apparatus for natural language parsing using multiple passes and tags
US7181438B1 (en) * 1999-07-21 2007-02-20 Alberti Anemometer, Llc Database access system
US7096502B1 (en) * 2000-02-08 2006-08-22 Harris Corporation System and method for assessing the security posture of a network
US6816878B1 (en) * 2000-02-11 2004-11-09 Steven L. Zimmers Alert notification system
CA2423965A1 (en) * 2000-09-29 2002-04-04 Gavagai Technology Incorporated A method and system for adapting synonym resources to specific domains
US7191252B2 (en) * 2000-11-13 2007-03-13 Digital Doors, Inc. Data security system and method adjunct to e-mail, browser or telecom program
US6829611B2 (en) * 2000-12-29 2004-12-07 Bellsouth Intellectual Property Corporation Data loader application
JP4171224B2 (en) * 2002-02-19 2008-10-22 株式会社日本総合研究所 Risk diagnosis system, risk diagnosis method and risk diagnosis program
US7373666B2 (en) * 2002-07-01 2008-05-13 Microsoft Corporation Distributed threat management
US7248159B2 (en) * 2003-03-01 2007-07-24 User-Centric Ip, Lp User-centric event reporting
US7999857B2 (en) * 2003-07-25 2011-08-16 Stresscam Operations and Systems Ltd. Voice, lip-reading, face and emotion stress analysis, fuzzy logic intelligent camera system
GB2407657B (en) * 2003-10-30 2006-08-23 Vox Generation Ltd Automated grammar generator (AGG)
US7813916B2 (en) * 2003-11-18 2010-10-12 University Of Utah Acquisition and application of contextual role knowledge for coreference resolution
US7464149B2 (en) * 2004-04-30 2008-12-09 International Business Machines Corporation System and method for managing introspectable objects in an enterprise
US7299152B1 (en) * 2004-10-04 2007-11-20 United States Of America As Represented By The Secretary Of The Navy Correlating event data for large geographic area
US8032594B2 (en) * 2004-11-10 2011-10-04 Digital Envoy, Inc. Email anti-phishing inspector

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6529934B1 (en) * 1998-05-06 2003-03-04 Kabushiki Kaisha Toshiba Information processing system and method for same
US20030023573A1 (en) * 2001-07-27 2003-01-30 International Business Machines Corporation Conflict-handling assimilator service for exchange of rules with merging

Also Published As

Publication number Publication date
US20130060772A1 (en) 2013-03-07
US20110016078A1 (en) 2011-01-20
WO2006076398B1 (en) 2008-01-10
US7822747B2 (en) 2010-10-26
US20060184483A1 (en) 2006-08-17
WO2006076398A3 (en) 2007-11-22
US8301628B2 (en) 2012-10-30

Similar Documents

Publication Publication Date Title
US8301628B2 (en) Predictive analytic method and apparatus
US11093568B2 (en) Systems and methods for content management
Faliagka et al. On-line consistent ranking on e-recruitment: seeking the truth behind a well-formed CV
US20160196313A1 (en) Personalized Question and Answer System Output Based on Personality Traits
US10783179B2 (en) Automated article summarization, visualization and analysis using cognitive services
Kaza et al. Evaluating ontology mapping techniques: An experiment in public safety information sharing
CN114341865B (en) Progressive concatenation for real-time conversations
Galitsky et al. Detecting logical argumentation in text via communicative discourse tree
Chan et al. Question-answering dialogue system for emergency operations
Elleuch et al. Discovering activities from emails based on pattern discovery approach
Jia et al. Pattern discovery and anomaly detection via knowledge graph
Derbas et al. Eventfully safapp: hybrid approach to event detection for social media mining
US20130054646A1 (en) Work optimization
Siahaan et al. User story extraction from natural language for requirements elicitation: Identify software-related information from online news
US20210390553A1 (en) Method for recommending and implementing communication optimizations
Huttunen et al. Predicting relevance of event extraction for the end user
WO2022234273A1 (en) Project data processing method and apparatus
Burstein et al. Decision support via text mining
Chang et al. Towards an automatic approach for assessing program competencies
Li et al. A rule-based Chinese sentiment mining system with self-expanding dictionary-taking TripAdvisor as an example
Rodger et al. Assessing American Presidential Candidates Using Principles of Ontological Engineering, Word Sense Disambiguation, and Data Envelope Analysis
Wu et al. NLP-based approach for automated safety requirements information retrieval from project documents
Lytvyn et al. COLINS 2019. Volume II: Workshop
Valiyev et al. Initial exploitation of natural language processing techniques on nato strategy and policies
Arini A conceptual model of trust in emergency evacuation: evidence from Indonesian volcano eruptions

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 06718031

Country of ref document: EP

Kind code of ref document: A2