Patent Application for Automatic Question and Answer Generation through Pattern Based Analysis and Natural Language Processing

Strategic Insights -

Researchers from International Business Machines Corporation have filed a patent application for a system that automatically generates test/training questions and answers through pattern based analysis and natural language processing techniques on a given corpus for quick domain adaptation.

The proposed system aims to alleviate the information gap that arises when users try to piece together relevant information from various structured and unstructured sources. The system would analyze an input question, decompose it into constituent parts, generate hypotheses based on the decomposed question and primary search results, perform hypothesis and evidence scoring, and finally output an answer to the input question along with a confidence measure.

Key Takeaways:

  • The patent application, filed on December 23, 2013, and made available online on July 2, 2015, describes a system that automatically generates question and answer pairs for training a question answering system for a given domain.
  • The system identifies a set of patterns of components in passages within a corpus of documents for the given domain, and applies a set of rules to the passages to generate the question and answer pairs.
  • The system utilizes natural language processing techniques to analyze the input question, decompose it into constituent parts, and generate hypotheses based on the decomposed question and primary search results.
  • The proposed system is designed to quickly adapt to new domains and improve the accuracy of question answering systems.
  • The inventors, Naveen G. Balani, Amit P. Bohra, and Krishna Kummamuru, are affiliated with International Business Machines Corporation.
  • The Watson.TM. system, developed by IBM, is an example of a question answering system that utilizes advanced natural language processing, information retrieval, knowledge representation and reasoning, and machine learning technologies.
  • The patent application cites U.S. Patent Application Publication Nos. 2011/0125734 and 2011/0066587, which describe prior art related to question and answer systems.

Statistics:

  • December 23, 2013: The patent application was filed.
  • July 2, 2015: The patent application was made available online.
  • 2011: The United States patent application Publications Nos. 2011/0125734 and 2011/0066587 were published.

Sources:

  • Balani, Naveen G.; Bohra, Amit P.; Kummamuru, Krishna. Automatically Generating Test/Training Questions and Answers through Pattern Based Analysis and Natural Language Processing Techniques on the Given Corpus for Quick Domain Adaptation. Filed December 23, 2013 and posted July 2, 2015.
  • Patent Application Publication No. 2011/0125734: U.S. Patent Application Publication No. 2011/0125734 discloses a mechanism for generating question and answer pairs based on a corpus of data.
  • Patent Application Publication No. 2011/0066587: U.S. Patent Application Publication No. 2011/0066587 discloses a mechanism for converting a report of analyzed information into a collection of questions and determining whether answers for the collection of questions are answered or refuted from the information set.