System and method for facts extraction and domain knowledge repository creation from unstructured and semi-structured documents

• US 8,423,495 B1
• Filed: 07/11/2012
• Issued: 04/16/2013
• Est. Priority Date: 06/18/2004
• Status: Active Grant

• Current Assignee: Glenbrook Networks
• Original Assignee: Glenbrook Networks
• Inventors: Komissarchik, Julia, Komissarchik, Edward
• Primary Examiner(s): Holmes, Michael B
• Application Number:
  US13/546,960
• Time in Patent Office: 279 Days
• Field of Search: 706/46
• US Class Current: 706/46
• CPC Class Codes: G06F 16/345 Summarisation for human users

Abstract

Provided are methods and systems that extract facts of unstructured documents and build an oracle for various domains. The present invention addresses the problem of efficient finding and extraction of facts about a particular subject domain from semi-structured and unstructured documents, makes inferences of new facts from the extracted facts and the ways of verification of the facts, thus becoming a source of knowledge about the domain to be effectively queried. The methods and systems can also extract temporal information from unstructured and semi-structured documents, and can find and extract dynamically generated documents from Deep or Dynamic Web.

Claims

1. A method for objects identification and inference comprising:

utilizing a three level object presentation consisting of instance, denotatum and denotatum class;

applying application dependent inference rules to determine a match between instances and objects.

2. The method of claim 1, wherein the mappings between instances and denotata are established within individual documents.

3. The method of claim 2 wherein the instance that can be mapped to more than one denotatum are excluded.

4. The method of claim 3, further comprising application-dependent rules are applied to establish the denotata from several pages as belonging to the same denotatum class, and thus representing one object.

5. A method for incorrect object identification recovery comprising:

utilizing a contradiction between new and old facts or human request as a trigger for reclassification of affected facts; and

utilizing the roll forward transactions to eliminate and rearrange denotatum classes.

6. The method of claim 5, wherein for all objects participated in affected facts the association of their denotata with denotatum classes is eliminated.

7. The method of claim 6, wherein all affected denotatum classes are deleted.

8. The method of claim 7, wherein all applicable inference rules are used to build a new set of denotatum classes and their association with affected denotata.

13. The method of claim 8, wherein a method of building implicit social network is provided, comprising:

utilizing association of people to the same companies, organizations, schools to establish relationship between individuals;

utilizing participation of people in the same events to establish relationships between individuals.

14. The method of claim 13, wherein the mentioned associations and participations are established by the means of business information network database.

15. The method of claim 14, wherein strength of a relationship is established as a combination of different factors selected from at least one of, positions of two individuals in a company hierarchy where they worked together, a number of times both individuals were participating in different events and level of their participation, and size of the school/class they attended together.

16. The method of claim 14, wherein the temporal boundaries of a particular relationship are established using time stamps of documents used to establish the relationship.

9. A method to convert unstructured and semi-structured information into a structured format, comprising:

crawling the Internet and Intranets to generate a set of pages for further analysis;

applying different knowledge agents in different order to each page to extract application dependent candidate facts;

building new candidate facts from the extracted facts using logical inference;

verifying correctness of the candidates facts using recursive verification, recursive bootstrapping and deferred decision methods; and

storing the verified facts in structured form in data repository;

wherein a method of building business information network database is provided, comprising;

collecting documents from internet and other sources;

applying surface and deep web crawling to collect these documents;

extracting business information facts from each document;

filtering out incorrect or irrelevant facts;

applying consistency checks, directly and recursively, to solidify correctness of facts;

storing facts in business information network database;

providing access to the information in business information network database for different on-line users.

10. The method of claim 9, wherein the business information network includes at least companies/organizations objects and people objects.

11. The method of claim 10, wherein the objects have relationships selected from at least one of, employment:

employee-employer, employer-employee;

business association;

vendor-customer, customer-vendor, partner-partner; and

person to person association;

worked together, studied together.

12. The method of claim 11, wherein objects and their relationships are presented as a hyper-graph, where nodes are objects, and relationships are hyperedges.

17. The method of claim 9, further comprising, utilizing a former association of people with former customers of a company after they became customers;

and utilizing a current association of the same people with potential customers, to build a customer alumni network.

18. The method of claim 17, wherein the relationship of type vendor-buyer from business information network is used to build a list of the customer'"'"'s employees, which had direct exposure to the company'"'"'s products or services.

19. The method of claim 18, wherein other employees from the customer, which belong to the immediate neighborhood of social network of the people on the list, are added to the list.

20. The method of claim 19, wherein individuals on the list cross referenced with the current employees of potential customers.

21. The method of claim 20, wherein individuals from the list that are currently working for potential customers are defined as members of customer alumni network.