Extraction of concept-based summaries from documents

Patent metadata

Field	Value
Publication number	US-9916375-B2
Application number	US-201414460558-A
Country	US
Kind code	B2
Filing date	Aug 15, 2014
Priority date	Aug 15, 2014
Publication date	Mar 13, 2018
Grant date	Mar 13, 2018

Abstract

Embodiments of the present invention enable users to generate a summary for a document with respect to a concept, making use of inherent hierarchies present in a text document based on subject-object relationships of the sentences in the text document. In one embodiment, a text document is parsed into sentences, and a tuple is created for each sentence, the tuple comprising a subject and an object found in the sentence. The tuples may then be searched for a specified topic to identify matching tuples, as well as tuples that are related to the matching tuples based on relationships between their respective subjects and objects. A summary focused on the specified topic may then be generated using the sentences corresponding to the matching tuples and the tuples related to the matching tuples.

First claim

What is claimed is:

1. A method for generating a summary of a document, the method comprising:
extracting, by one or more computer processors, text content from the document;
parsing, by one or more computer processors, the text content into a plurality of sentences;
parsing, by one or more computer processors, each sentence of the plurality of sentences into a subject and an object;
for each sentence of the plurality of sentences, creating, by one or more computer processors, a tuple corresponding to the sentence, the tuple comprising a subject of the sentence and an object of the sentence;
receiving, by one or more computer processors, a topic about which to generate the summary;
searching, by one or more computer processors, the plurality of created tuples to identify one or more tuples that match the received topic, wherein searching the plurality of created tuples to identify one or more tuples that match the received topic comprises:
identifying, by one or more computer processors, a first tuple of the plurality of created tuples that has a subject that matches the topic;
identifying, by one or more computer processors, a second tuple of the plurality of created tuples that has an object that matches the topic;
identifying, by one or more computer processors, one or more tuples in one or more levels of a first hierarchy, wherein at least one tuple in one level of the first hierarchy has an object that matches the subject of the first tuple, and each tuple in each other level of the first hierarchy has an object that matches a subject of all tuples in another level of the first hierarchy; and
identifying, by one or more computer processors, one or more tuples in one or more levels of a second hierarchy, wherein at least one tuple in one level of the second hierarchy has a subject that matches the object of the second tuple, and each tuple in each other level of the second hierarchy has a subject that matches an object of all tuples in another level of the second hierarchy; and

generating, by one or more computer processors, a summary of the document comprising the sentences that correspond to the one or more identified tuples that match the topic.

2. The method of claim 1 , wherein searching the plurality of created tuples to identify one or more tuples that match the received topic further comprises:
identifying, by one or more computer processors, a third tuple of the plurality of created tuples, the third tuple having a subject that matches the object of the first tuple; and
identifying, by one or more computer processors, a fourth tuple of the plurality of created tuples, the fourth tuple having an object that matches the subject of the second tuple.

3. The method of claim 1 , wherein generating, by one or more computer processors, a summary of the document comprising the sentences that correspond to the one or more identified tuples that match the topic comprises:
generating, by one or more computer processors, a summary comprising sentences corresponding to the first tuple and the second tuple of the plurality of created tuples.

4. The method of claim 2 , wherein generating, by one or more computer processors, a summary of the document comprising the sentences that correspond to the one or more identified tuples that match the topic comprises:
generating, by one or more computer processors, a summary comprising sentences corresponding to the first tuple, the second tuple, the third tuple, and the fourth tuple of the plurality of created tuples.

5. The method of claim 1 , wherein an amount of levels in the one or more levels in the first and second hierarchies is controlled by one or more parameters.

6. A computer program product for generating a summary of a document, the computer program product comprising:
one or more computer readable storage media and program instructions stored on the one or more computer readable storage media, the program instructions comprising:
program instructions to extract text content from a document;
program instructions to parse the text content into a plurality of sentences;
program instructions to parse each sentence of the plurality of sentences into a subject and an object;
program instructions to, for each sentence of the plurality of sentences, create a tuple corresponding to the sentence, the tuple comprising a subject of the sentence and an object of the sentence;
program instructions to receive a topic about which to generate the summary;
program instructions to search the plurality of created tuples to identify one or more tuples that match the received topic, wherein the program instructions to search the plurality of created tuples to identify one or more tuples that match the received topic comprise:
program instructions to identify a first tuple of the plurality of created tuples that has a subject that matches the topic;
program instructions to identify a second tuple of the plurality of created tuples that has an object that matches the topic;
program instructions to identify one or more tuples in one or more levels of a first hierarchy, wherein at least one tuple in one level of the first hierarchy has an object that matches the subject of the first tuple, and each tuple in each other level of the first hierarchy has an object that matches a subject of all tuples in another level of the first hierarchy; and
program instructions to identify one or more tuples in one or more levels of a second hierarchy, wherein at least one tuple in one level of the second hierarchy has a subject that matches the object of the second tuple, and each tuple in each other level of the second hierarchy has a subject that matches an object of all tuples in another level of the second hierarchy; and

program instructions to generate a summary of the document comprising the sentences that correspond to the one or more identified tuples that match the topic.

7. The computer program product of claim 6 , wherein the program instructions to search the plurality of created tuples to identify one or more tuples that match the received topic further comprise:
program instructions to identify a third tuple of the plurality of created tuples, the third tuple having a subject that matches the object of the first tuple; and
program instructions to identify a fourth tuple of the plurality of created tuples, the fourth tuple having an object that matches the subject of the second tuple.

8. The computer program product of claim 6 , wherein the program instructions to generate a summary of the document comprising the sentences that correspond to the one or more identified tuples that match the topic comprise:
program instructions to generate a summary comprising sentences corresponding to the first tuple and the second tuple of the plurality of created tuples.

9. The computer program product of claim 7 , wherein the program instructions to generate a summary of the document comprising the sentences that correspond to the one or more identified tuples that match the topic comprise:
program instructions to generate a summary comprising sentences corresponding to the first tuple, the second tuple, the third tuple, and the fourth tuple of the plurality of created tuples.

10. The computer program product of claim 6 , wherein an amount of levels in the one or more levels in the first and second hierarchies is controlled by one or more parameters.

11. A computer system for generating a summary of a document, the computer system comprising:
one or more computer processors;
one or more computer readable storage media;
program instructions stored on the computer readable storage media for execution by at least one of the one or more processors,

Assignees

• IBM

Inventors

• Bhagwat Anand B
• Kummamuru Krishna

Classifications

• G06F16/313 Primary

  Selection or weighting of terms for indexing · CPC title

• G06F16/345

  Summarisation for human users · CPC title

• G06F17/30719

  Physics · mapped topic

• G06F17/30616 Primary

  Physics · mapped topic