SIGNIFICANCE OF RELATIONSHIPS DISCOVERED IN A CORPUS

• US 20170161263A1
• Filed: 05/25/2016
• Published: 06/08/2017
• Est. Priority Date: 12/02/2015
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Gordon, John B., Hogan, John P., Kottaram, Sanjay F.
• Granted Patent: US 9,916,536 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 40/40 Processing or translation of natural language natural language analysis G06F40/20; semantic analysis G06F40/30
• CPC Class Codes: G06F 9/453 Help systems
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/02 Neural networks
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G09B 7/04 characterised by modifying the teaching programme in response to a wrong answer, e.g. repeating the question, supplying a further explanation

Abstract

Certain relationships representing material insights are identified from among a set of discovered relationships. Cognitive discovery of relationships in a knowledge base, or corpus, are ranked according to one or more metrics indicative of material insights, including recentness and degree of alignment.

Claims

1. A method comprising:

ingesting a first body of information for a domain of knowledge;

receiving a natural language question corresponding to the domain of knowledge, the natural language question being in the form of a complete human language question;

parsing the natural language question to identify a focus;

mining the body of information for entities and relationships among entities based on the focus;

generating a targeting document for the domain of knowledge containing a first relationship and a second relationship, the first and second relationships being among entities mined from the first body of information;

recording mining data for relationships in the targeting document in support of an importance criteria;

ranking the first relationship with respect to the second relationship according to the importance criteria as a set of ranking data; and

storing the targeting document including the set of ranking data in the first body of information for on-demand access during a question-answer session corresponding to the domain of knowledge.

2. The method of claim 1, wherein the importance criteria is based upon an age of the first relationship compared to an age of the second relationship.

3. The method of claim 1, wherein the importance criteria relates to a comparison between a first set of evidence that identifies the first relationship with a second set of evidence that identifies the second relationship.

4. The method of claim 3, wherein the comparison between the first set of evidence and the second set of evidence is based at least in part upon a number of evidentiary documents within each of the first set of evidence and the second set of evidence.

5. The method of claim 9, further comprising:

incorporating the new document into the first body of information to form a second body of information.

6. The method of claim 1, further comprising:

registering the corresponding set of relationships in a relationship store having a set of relationships identified from the first body of information.

7. The method of claim 1 further comprising:

reporting to a user a set of top-ranked relationships in the targeting document.

8. The method of claim 1, further comprising:

classifying an identified relationship as a less-trusted fact; and

modifying the classification of the identified relationship from the less-trusted fact to a more-trusted fact based at least in part upon a count of trusted sources that support the identified relationship.

9. The method of claim 1, wherein the natural language question is presented as a training question corresponding to a new document to be added to the first body of information.

10. The method of claim 1, further comprising:

reporting the set of ranking data as input to determine a restated question.

11. The method of claim 1, wherein the importance criteria is a degree to which first relationship is known and a degree to which the second relationship is known according to a dictionary of commonly known relationships for the domain of knowledge.

12. The method of claim 4, wherein the comparison relates to a first count of evidentiary documents within the first set of evidence and a second count of evidentiary documents within the second set of evidence, the first and second counts representing the number of evidentiary documents available outside of the domain of knowledge.