Context based passage retreival and scoring in a question answering system

Patent metadata

Field	Value
Publication number	US-9529894-B2
Application number	US-201615044161-A
Country	US
Kind code	B2
Filing date	Feb 16, 2016
Priority date	Nov 7, 2014
Publication date	Dec 27, 2016
Grant date	Dec 27, 2016

Abstract

According to an aspect, a question answering computer system determines a set of search terms for a given question. A plurality of passages that each includes at least one search term in the set of search terms is identified in response to a search of a document for search terms in the set of search terms. A passage from the plurality of passages is selected. A context based score of the selected passage that indicates the likelihood that the selected passage contains an answer to the question is generated based on criteria that specify a relationship among the set of search terms, content of the passage, and content of at least one other passage in the plurality of passages.

First claim

What is claimed is:

1. A method comprising:
determining, by a question answering computer system, a set of search terms for a given question;
identifying, by the question answering computer system, a plurality of passages in response to a search of a document for search terms in the set of search terms, wherein each of the plurality of passages includes at least one search term in the set of search terms, and the document contains unstructured content that is organized in one of a narrative form and a hierarchy form;
selecting, by the question answering computer system, a passage from the plurality of passages; and
generating, by the question answering computer system, a context based score of the selected passage that indicates a likelihood that the selected passage contains an answer to the given question, the generating based on criteria that specify a relationship among the set of search terms, content of the selected passage, and content of an at least one other passage in the plurality of passages that precedes the selected passage in the document, comprising:
calculating a first score based on the content of the selected passage and a first subset of the set of search terms, the content of the selected passage including the first subset;
creating a list containing a second subset of the set of search terms, the second subset is different than the first subset and is not included in the content of the selected passage;
performing repeatedly until the list is empty:
determining whether there is at least one other passage in the plurality of passages that precedes the selected passage in the document;
selecting one of the at least one other passage in the plurality of passages that precedes the selected passage in the document in response to determining that there is at least one other passage in the plurality of passages that precedes the selected passage in the document, the one of the at least one other passage in the plurality of passages that precedes the selected passage in the document contains at least one search term in the list, the selecting is performed in an order of closest proximity to the selected passage in the document;
generating a second score by scoring the one of the at least one other passage that precedes the selected passage in the document based on the at least one search term in the list that is contained in the one of the at least one other passage in the plurality of passages that precedes the selected passage in the document, and removal of the first subset from the at least one other passage that precedes the selected passage in the document:
updating the second score based on a distance in the document between the selected passage and the one of the at least one other passage in the plurality of passages that precedes the selected passage in the document;
aggregating the first score and the updated second score by adding the first score to the updated second score to generate the context based score; and
in response to the adding, updating the list by removing from the list all search terms, including the at least one search term contained in the at least one other passage in the plurality of passages the precedes the selected passage in the document.

Assignees

• IBM

Inventors

• Nolan Christopher M
• Visotski William G
• Wilson David E
• Yates Robert L

Classifications

• G06F16/334 Primary

  Query execution (filtering based on additional data G06F16/335) · CPC title

• G06F16/24578

  using ranking · CPC title

• G06F16/90344

  by using string matching techniques · CPC title

• G06F16/335

  Filtering based on additional data, e.g. user or group profiles (filtering in web context G06F16/9535, G06F16/9536) · CPC title

• G06N5/00

  Computing arrangements using knowledge-based models · CPC title