Automatic, unsupervised paraphrase detection

• US 10,275,453 B2
• Filed: 06/06/2017
• Issued: 04/30/2019
• Est. Priority Date: 05/12/2017
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Bennett, Laura J., Krishnamurthy, Lakshminarayanan, Parameswaran, Niyati, Sudarsan, Sridhar
• Primary Examiner(s): Singh, Satwant K
• Application Number:
  US15/615,189
• Publication Number:
  US 20180329882A1
• Time in Patent Office: 693 Days
• Field of Search: None
• CPC Class Codes: G06F 16/243 Natural language query formulation
• CPC Class Codes: G06F 40/20 Natural language analysis semantic analysis of natural language G06F40/30
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/041 Abduction

Abstract

A method is disclosed for identifying paraphrases in a natural language processing (NLP) system comprising: receiving a first phrase and a second phrase by a system; analyzing the first phrase and the second phrase to provide a semantic and structural hierarchical comparison assessment, the semantic and structural hierarchical comparison assessment having an associated semantic and structural hierarchical comparison assessment value; and determining whether the semantic and structural hierarchical comparison assessment value exceeds a predetermined paraphrase equivalency criteria; and, responsive to determining the semantic and structural hierarchical comparison assessment value exceeds the predetermined paraphrase equivalency criteria, classifying the second phrase as being a rewording of the first phrase.

Claims

1. A computer-implementable method for identifying paraphrases in a natural language processing (NLP) system comprising:

receiving a first phrase and a second phrase by a system;

analyzing the first phrase and the second phrase to provide a semantic and structural hierarchical comparison assessment, the semantic and structural hierarchical comparison assessment having an associated semantic and structural hierarchical comparison assessment value; and

determining whether the semantic and structural hierarchical comparison assessment value exceeds a predetermined paraphrase equivalency criteria;

responsive to determining the semantic and structural hierarchical comparison assessment value exceeds the predetermined paraphrase equivalency criteria, classifying the second phrase as being a rewording of the first phrase.

2. The method of claim 1, wherein:

the structural hierarchical comparison assessment includes a vector based and a semantic based framework.

3. The method of claim 2, wherein:

the first phrase and the second phrase are in a single language.

4. The method of claim 3, further comprising:

utilizing the first phrase and the second phrase in a ground truth classifier for a system capable of answering questions.

5. The method of claim 1, wherein:

the semantic and structural hierarchical comparison assessment comprises performing an ensemble operation on the first phrase and the second phrase, the ensemble operation providing a structural similarity score.

6. The method of claim 1, wherein:

the semantic and structural hierarchical comparison assessment comprises performing a semantic similarity operation, the semantic similarity operation providing a semantic similarity score.