System, method, and computer program product for combination of cognitive causal models with reasoning and text processing for knowledge driven decision support

• US 7,644,053 B2
• Filed: 03/02/2005
• Issued: 01/05/2010
• Est. Priority Date: 03/03/2004
• Status: Active Grant

• Current Assignee: The Boeing Co.
• Original Assignee: The Boeing Co.
• Inventors: Harrison, Philip, Kipersztok, Oscar
• Primary Examiner(s): Vincent; David R
• Assistant Examiner(s): Wong; Lut
• Application Number:
  US11/070,452
• Publication Number:
  US 20050197992A1
• Time in Patent Office: 1,770 Days
• Field of Search: 706/46, 706/50
• US Class Current: 706/46
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

A system, method, and computer program product for combining causal domain models with reasoning and text processing for knowledge driven decision support are provided. A knowledge driven decision support system is capable of creating a domain model, extracting and processing quantities of text according to the domain model, and generating understanding of the content and implications of information sensitive to analysts. An interface may be used to receive input to model complex relationships of a domain, establish implications of interest or request a query, and update the causal model. A processing element can capture and process text into text profiles by incorporating the domain model and process the text profiles in accordance with the domain model by applying formal reasoning to the information to derive trends, predict events, or arrive at other query results. An output element can provide a user the resulting information related to the domain model.

Claims

1. A system for assisting knowledge driven decision support, comprising:

a processor configured to operate a directed acyclic graph Bayesian network of an unconstrained causal domain model, a text processing tool, and a reasoning tool;

the unconstrained causal domain model configured to define and store, on a computer-readable medium, at least two domain concepts and at least one causal relationship between the domain concepts, wherein the directed acyclic graph Bayesian network is configured to represent a formalization of the unconstrained causal domain model with minimal information loss by eliminating cycles of the unconstrained causal domain model by computing information gain and eliminating influence arcs of the unconstrained causal domain model that minimizes information loss;

the text processing tool configured to analyze at least one document using the directed acyclic graph Bayesian network to produce at least one text profile for the at least one document; and

the reasoning tool configured to determine a result by analyzing the at least one text profile using at least one domain concept of the directed acyclic graph Bayesian network and one causal relationship of the domain concept, wherein the reasoning tool is further configured to use the text processing tool to use the directed acyclic graph Bayesian network, and wherein the reasoning tool is further configured to electronically present a display identifying the result of the analysis of the reasoning tool, providing a user the ability to use the result in a decision making process.

2. The system of claim 1, wherein the text processing tool comprises:

a text acquisition tool configured to acquire and store, on a computer-readable medium, text from at least one document.

3. The system of claim 2, wherein the text acquisition tool comprises:

a data mining tool configured to mine at least one document;

a document retrieval tool configured to receive at least one mined document; and

a text extraction tool configured to extract text from at least one retrieved document to provide text.

4. The system of claim 2, wherein the text acquisition tool further comprises:

a text profiling tool configured to profile the acquired text, wherein the text profiling tool is further configured to use at least one domain concept of the unconstrained causal domain model to create and store, on a computer-readable medium, text profiles of the text.

5. The system of claim 1, wherein the reasoning tool comprises:

a transformation routine configured to transform the unconstrained causal domain model into a mathematical formalization of the unconstrained causal domain model.

6. The system of claim 1, wherein the reasoning tool comprises a Bayesian network belief update algorithm.

7. The system of claim 1, further comprising:

a common sense knowledge base configured to be used by at least one of the text processing tool and the reasoning tool.

8. The system of claim 1, further comprising:

an interface configured to receive input for creating the unconstrained causal domain model.

9. The system of claim 1, further comprising:

an interface configured to receive input for creating the unconstrained causal domain model;

wherein the processor is further configured to transform the unconstrained causal domain model into a mathematical formalization of the unconstrained causal domain model, communicate with the text processing tool for analyzing text of documents in accordance with the unconstrained causal domain model to create text profiles, and communicate with the reasoning tool for performing reasoning analysis upon the text profiles in accordance with the unconstrained causal domain model using the direct acyclic graph Bayesian network to derive a result; and

an output element configured to present the result.

10. The system of claim 9, wherein the unconstrained causal domain model comprises a plurality of domain concept nodes associated with domain concepts and interconnected by relationship arcs associated with causal relationships between the domain concepts in a manner defined by at least one of the input relating to the creation of the unconstrained causal domain model, the processing of text, and the performance of reasoning analysis.

11. A method, comprising:

providing an unconstrained causal domain model defined by domain concepts and causal relationships between the domain concepts;

transforming, using a processor, the unconstrained causal domain model into a directed acyclic graph Bayesian network formalism with minimal information loss, wherein the transformation comprises eliminating cycles of the unconstrained causal domain model by computing information gain and eliminating influence arcs of the unconstrained causal domain model that minimizes information loss;

storing on a computer-readable medium, the directed acyclic graph Bayesian network formalism;

analyzing text using the direct acyclic graph Bayesian network formalism to extract information and derive test profiles;

storing, on a computer-readable medium, the text profiles;

performing reasoning analysis of the text profiles according to the directed acyclic graph Bayesian network formalism to derive a result; and

electronically presenting a display identifying the result of the reasoning analysis, providing a user the ability to use the result in a decision making process.

12. The method of claim 11, further comprising:

creating the unconstrained causal domain model by defining the domain concepts and the causal relationships, wherein at least one of the domain concepts and the causal relationships are configured to be used to derive text profiles, formalize the unconstrained causal domain model as the directed acyclic graph Bayesian network, and perform reasoning analysis; and

storing, on a computer-readable medium, the unconstrained causal domain model.

13. The method of claim 11, wherein performing reasoning analysis comprises performing a Bayesian network belief update algorithm.

14. The method of claim 11, wherein performing reasoning analysis comprises performing a predictive analysis using an inference algorithm.

15. The method of claim 11, further comprising:

modifying the unconstrained causal domain model to define a query.

16. The method of claim 11, further comprising:

validating the unconstrained causal domain model.

17. The method of claim 11, further comprising:

updating the unconstrained causal domain model.

18. The method of claim 11, further comprising acquiring the text for analyzing using the directed acyclic graph Bayesian network formalism, wherein acquiring the text comprises:

mining external sources for documents;

retrieving the mined documents from the external sources; and

extracting the text from the retrieved documents.

19. The method of claim 18, further comprising processing the text.

20. The method of claim 19, wherein processing the text comprises:

relevance filtering the text for relevance to the unconstrained causal domain model; and

event filtering the text for events associated with the unconstrained causal domain model.

21. The method of claim 19, wherein processing the text uses a common sense knowledge base, and wherein the common sense knowledge base is generated by:

extracting knowledge from a corpus of text, wherein the extracted knowledge is transformed into formal logic representing the extracted knowledge; and

manually capturing knowledge, wherein the captured knowledge is transformed into formal logic representing the captured knowledge.

22. A computer-readable medium encoded with a computer program for causing a processor to support a decision using an unconstrained causal domain model, the computer program comprising:

a first code for causing the processor to provide the unconstrained causal domain model;

a second code for causing the processor to transform the unconstrained causal domain model into a directed acyclic graph Bayesian network formalism with minimal information loss by eliminating cycles of the unconstrained causal domain model by computing information gain and eliminating influence arcs of the unconstrained causal domain model that minimizes information loss and store, on a computer-readable medium, the directed acyclic graph Bayesian network formalism; and

a third code for causing the processor to analyze text using the directed acyclic graph Bayesian network formalism to derive text profiles and store, on a computer-readable medium, the text profiles;

a fourth code for causing the processor to perform reasoning analysis of the text profiles according to the directed acyclic graph Bayesian network formalism.

23. The computer-readable medium of claim 22, wherein the computer program further comprises:

a fifth code for causing the processor to modify the unconstrained causal domain model to define a query.

24. The computer-readable medium of claim 22, wherein the computer program further comprises:

a sixth code for causing the processor to validate the unconstrained causal domain model.

25. The computer-readable medium of claim 22, wherein the computer program further comprises:

a seventh code for causing the processor to acquire the text for analyzing using the directed acyclic graph Bayesian network formalism.

26. The computer-readable medium of claim 25, wherein the seventh code comprises:

an eighth code for causing the processor to mine external sources for documents;

a ninth code for causing the processor to retrieve the mined documents from the external sources; and

a tenth code for causing the processor to acquire the text from the retrieved documents.

27. The computer-readable medium of claim 25, wherein the seventh code comprises:

an eleventh code for causing the processor to process the text to derive the processed text.

28. The computer-readable medium of claim 27, wherein the eleventh code comprises:

a twelfth code for causing the processor to filter the text for relevance to the unconstrained causal domain model; and

a thirteenth code for causing the processor to filter the text for events associated with the unconstrained causal domain model.

29. The computer-readable medium of claim 22, wherein the fourth code is further configured to derive a result, and further comprising a fifth code for causing the processor to electronically present a display identifying the result of the fourth code, providing a user the ability to use the result in a decision making process.