Claim generation

Patent metadata

Field	Value
Publication number	US-10776587-B2
Application number	US-201615206326-A
Country	US
Kind code	B2
Filing date	Jul 11, 2016
Priority date	Jul 11, 2016
Publication date	Sep 15, 2020
Grant date	Sep 15, 2020

Abstract

A computer-implemented method, computerized apparatus and computer program product for claim generation, the method comprising: selecting at least one subject according to a given topic; selecting at least one verb from a first data source; selecting at least one object from a second data source; generating one or more candidate claim sentences, each of which composed of a subject selected from the at least one subject, a verb selected from the at least one verb and an object selected from the at least on object; and determining validity of the candidate claim sentences using a machine learning process.

First claim

What is claimed is:

1. A computer-implemented method comprising:
selecting at least one subject according to a given topic;
selecting at least one verb from a first data source;
selecting at least one object from a second data source different from the first data source;
automatically synthesizing one or more automatically-generated candidate claim sentences, comprising combining for each automatically-generated candidate claim sentence: a subject selected from the at least one subject, a verb selected from the at least one verb and an object selected from the at least one object; and
determining validity of the candidate automatically-generated claim sentences using a machine learning process, comprising applying a trained classifier on one or more features extracted from a candidate claim sentence selected from the automatically-generated candidate claim sentences, to obtain a predicted validity labeling, the predicated validity labeling referring to the coherency and relevancy to the given topic.

2. The computer-implemented method of claim 1 , wherein said selecting at least one subject comprises analyzing the given topic to extract one or more noun phrases.

3. The computer-implemented method of claim 1 , wherein the first data source is selected from the group consisting of: verbs extracted from a set of pre-labeled claim sentences; predetermined causation verbs; and any combination thereof.

4. The computer-implemented method of claim 1 , wherein the second data source is selected from the group consisting of: objects extracted from a set of pre-labeled claim sentences; objects extracted from a set of claim sentences detected in a corpus; concepts related to the given topic in a knowledge base; objects mined from a corpus based on predetermined criteria; and any combination thereof.

5. The computer-implemented method of claim 1 , wherein the trained classifier is trained using a training dataset comprising pre-labeled claim sentences.

6. The computer-implemented method of claim 1 , wherein at least one of the features relates to similarity between constituents of the candidate claim sentence and the given topic.

7. The computer-implemented method of claim 1 , wherein at least one of the features relates to co-occurrence of constituents of the candidate claim sentence within a corpus.

8. The computer-implemented method of claim 1 , wherein the first and second data sources comprise a same set of pre-labeled claim sentences associated each with a topic, wherein a verb and an object of a candidate claim sentence are selected from a same pre-labeled claim sentence of the set, wherein at least one of the features relates to similarity between constituents of the pre-labeled claim sentence and associated topic thereof.

9. The computer-implemented method of claim 1 , further comprising: using the machine learning process to rank candidate claim sentences determined as valid, and outputting a top ranked subset thereof.

10. A computerized apparatus having a processor, the processor being adapted to perform the steps of:
selecting at least one subject according to a given topic;
selecting at least one verb from a first data source;
selecting at least one object from a second data source different from the first data source;
automatically synthesizing one or more automatically-generated candidate claim sentences, comprising combining for each automatically-generated candidate claim sentence: a subject selected from the at least one subject, a verb selected from the at least one verb and an object selected from the at least one object; and
determining validity of the candidate automatically-generated claim sentences using a machine learning process, comprising applying a trained classifier on one or more features extracted from a candidate claim sentence selected from the automatically-generated candidate claim sentences, to obtain a predicted validity labeling, the predicated validity labeling referring to the coherency and relevancy to the given topic.

11. The computerized apparatus of claim 10 , wherein said selecting at least one subject comprises analyzing the given topic to extract one or more noun phrases.

12. The computerized apparatus of claim 10 , wherein the first data source is selected from the group consisting of: verbs extracted from a set of pre-labeled claim sentences: predetermined causation verbs; and any combination thereof.

13. The computerized apparatus of claim 10 , wherein the second data source is selected from the group consisting of: objects extracted from a set of pre-labeled claim sentences; objects extracted from a set of claim sentences detected in a corpus; concepts related to the given topic in a knowledge base: objects mined from a corpus based on predetermined criteria; and any combination thereof.

14. The computerized apparatus of claim 10 , wherein at least one of the features relates to similarity between constituents of the candidate claim sentence and the given topic.

15. The computerized apparatus of claim 10 , wherein at least one of the features relates to co-occurrence of constituents of the candidate claim sentence within a corpus.

16. The computerized apparatus of claim 10 , wherein the first and second data sources comprise a same set of pre-labeled claim sentences associated each with a topic, wherein a verb and an object of a candidate claim sentence are selected from a same pre-labeled claim sentence of the set, wherein at least one of the features relates to similarity between constituents of the pre-labeled claim sentence and associated topic thereof.

17. The computerized apparatus of claim 10 , wherein said processor is further configured for using the machine learning process to rank candidate claim sentences determined as valid, and outputting a top ranked subset thereof.

18. A computer program product comprising a non-transitory computer readable storage medium retaining program instructions, which program instructions when read by a processor, cause the processor to perform a method comprising:
selecting at least one subject according to a given topic;
selecting at least one verb from a first data source;
selecting at least one object from a second data source different from the first data source;
automatically synthesizing one or more automatically-generated candidate claim sentences, comprising combining for each automatically-generated candidate claim sentence: a subject selected from the at least one subject, a verb selected from the at least one verb and an object selected from the at least one object; and
determining validity of the candidate automatically-generated claim sentences using a machine learning process, comprising applying a trained classifier on one or more features extracted from a candidate claim sentence selected from the automatically-generated candidate claim sentences, to obtain a predicted validity labeling, the predicated validity labeling referring to the coherency and relevancy to the given topic.

Assignees

• IBM

Inventors

• Bilu Yonatan
• Levy Ran
• Slonim Noam

Classifications

• G06N20/00

  Machine learning · CPC title

• G06F40/35 Primary

  Discourse or dialogue representation · CPC title

• G06F40/56 Primary

  Natural language generation · CPC title