Relationship extraction

Patent metadata

Field	Value
Publication number	US-10643145-B2
Application number	US-201315027810-A
Country	US
Kind code	B2
Filing date	Nov 27, 2013
Priority date	Nov 27, 2013
Publication date	May 5, 2020
Grant date	May 5, 2020

Abstract

Relationship extraction can include applying unsupervised relationship learning to a logic knowledge base and a plurality of entity groups recognized from a document to provide a probabilistic model. Relationship extraction can include performing joint inference on the probabilistic model to make simultaneous statistical judgments about a respective relationship between at least two entities in one of the plurality of entity groups. Relationship extraction can include extracting a relationship between at least two entities in one of the plurality of entity groups based on the joint inference.

First claim

What is claimed:

1. A non-transitory machine-readable medium storing instructions for relationship extraction executable by a machine to cause the machine to:
apply unsupervised relationship learning to a logic knowledge base and a plurality of entities extracted from a document to generate a plurality of first-order logic formulas that represent explicit relationships between the plurality of entities as described in the document;
associate a plurality of weights to the plurality of first-order logic formulas;
subsequent to associating the plurality of weights to the plurality of first-order logic formulas, combine the plurality of first-order logic formulas to create a new first-order logic formula of a probabilistic model for inferring an implicit relationship between two entities of the plurality of entities, wherein the two entities have an undefined relationship in the document;
perform a joint inference on the new first-order logic formula of the probabilistic model to define the implicit relationship between the two entities; and
extract the implicit relationship between the two entities based on the joint inference.

2. The non-transitory machine-readable medium of claim 1 , wherein the implicit relationship between the two entities is a most likely relationship between the two entities.

3. The non-transitory machine-readable medium of claim 1 , wherein the probabilistic model includes the plurality of first-order logic formulas and the plurality of weights associated with the plurality of first-order logic formulas.

4. The non-transitory machine-readable medium of claim 3 , wherein each of the plurality of weights is associated with one of the plurality of first-order logic formulas.

5. The non-transitory machine-readable medium of claim 4 , wherein the plurality of weights collectively provide a plurality of probabilities that are associated with the plurality of first-order logic formulas.

6. The non-transitory machine-readable medium of claim 5 , wherein the plurality of probabilities associated with the plurality of first-order logic formulas are provided via a log-linear model.

7. The non-transitory machine-readable medium of claim 1 , wherein the instructions executable to cause the machine to extract the implicit relationship between the two entities include instructions that are executable to cause the machine to use the new first-order logic formula to extract the implicit relationship between the two entities.

8. A system for relationship extraction comprising a processing resource and a non-transitory machine readable medium storing instructions that when executed by the processing resource cause the processing resource to:
apply unsupervised relationship learning to a first-order logic knowledge base and a plurality of entities extracted from a document to generate a plurality of first-order logic formulas that represent explicit relationships between the plurality of entities as described in the document;
associate a plurality of weights to the plurality of first-order logic formulas;
subsequent to associating the plurality of weights to the plurality of first-order logic formulas, combine the plurality of first-order logic formulas to create a new first-order logic formula of a probabilistic graphical model for inferring an implicit relationship between two entities of the plurality of entities, wherein the two entities have an undefined relationship in the document;
perform a joint inference on the new first-order logic formula of the probabilistic graphical model to define the implicit relationship between the two entities; and
extract the implicit relationship between the two entities based on the joint inference.

9. The system of claim 8 , including instructions executable to cause the processing resource to extract a relationship between a recognized entity pair of the plurality of entities.

10. The system of claim 9 , including instructions executable to cause the processing resource to make a plurality of probabilistic determinations in parallel for a plurality of recognized entity pairs of the plurality of entities.

11. The system of claim 8 , wherein the instructions executable to cause the processing resource to extract the implicit relationship between the two entities includes instructions executable to cause the processing resource to relationally auto-correlate a variable pertaining to a first recognized entity pair with a variable pertaining to a second recognized entity pair to extract the implicit relationship between the two entities based on the joint inference.

12. The system of claim 8 , wherein the instructions executable to cause the processing resource to associate the plurality of weights to the plurality of first-order logic formulas include instructions that are executable to cause the processing resource to associate each of the plurality of weights to one of the plurality of first-order logic formulas.

13. The system of claim 12 , wherein the plurality of weights collectively provide probabilities associated with the plurality of first-order logic formulas, and wherein the probabilities associated with the plurality of first-order logic formulas are provided via a log-linear model.

14. A method for relationship extraction comprising:
applying, by a processor, unsupervised relationship learning to a first-order logic knowledge base and a plurality of entities extracted from a document to generate a plurality of first-order logic formulas that represent explicit relationships between the plurality of entities as described in the document;
associating, by the processor, a plurality of weights to the plurality of first-order logic formulas;
subsequent to associating the plurality of weights to the plurality of first-order logic formulas, combining the plurality of first-order logic formulas to create a new first-order logic formula of a probabilistic graphical model for inferring an implicit relationship between two entities of the plurality of entities, wherein the two entities have an undefined relationship in the document;
performing, by the processor, a joint inference on the new first-order logic formula of the probabilistic graphical model to define the implicit relationship between the two entities; and
extracting, by the processor, the implicit relationship between the two entities based on the joint inference.

15. The method of claim 14 , wherein the document is a textual document and does not provide explicit support for the implicit relationship of the two entities.

16. The method of claim 14 , wherein associating the plurality of weights to the plurality of first-order logic formulas includes associating each of the plurality of weights to one of the plurality of first-order logic formulas.

17. The method of claim 16 , wherein the plurality of weights collectively provide probabilities associated with the plurality of first-order logic formulas, and the probabilities associated with the plurality of first-order logic formulas are provided via a log-linear model.

18. The method of claim 14 , wherein the implicit relationship between the two entities is a most likely relationship between the two entities.

Assignees

• Hewlett Packard Entpr Dev Lp
• Yu Xiaofeng
• Micro Focus Llc

Inventors

• Yu Xiaofeng

Classifications

• G06F40/30

  Semantic analysis · CPC title

• G06F16/248

  Presentation of query results · CPC title

• G06F40/295

  Named entity recognition · CPC title

• G06N5/046

  Forward inferencing; Production systems · CPC title

• G06N7/005

  Physics · mapped topic