GENERATION OF SEARCH RESULT ABSTRACTS

• US 20100057710A1
• Filed: 08/28/2008
• Published: 03/04/2010
• Est. Priority Date: 08/28/2008
• Status: Active Grant

• Current Assignee: R2 Solutions LLC (Acacia Research Corporation)
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Pedersen, Jan, Kanungo, Tapas, Sarlos, Tamas
• Granted Patent: US 8,984,398 B2
• Time in Patent Office: Days
• US Class Current: N/A
• CPC Class Codes: G06F 16/345 Summarisation for human users

Abstract

The subject matter disclosed herein relates to generating an abstract of an electronic document based at least in part on a utility metric associated with a group phrases as well as on a ranked order of sentences.

Claims

1. A method comprising:

extracting a group of two or more sets of phrases represented within a computing platform by digital electronic signals from a set of two or more sentences associated with an electronic document, wherein a single sentence of said two or more sentences is associated with a single set of one or more phrases;

generating an abstract of said electronic document represented within a computing platform by digital electronic signals using at least one phrase from said group of two or more sets of phrases based at least in part on a utility metric associated with said group of two or more sets of phrases as well as on a ranked order of said two or more sentences; and

transmitting a search result, represented within a computing platform by digital electronic signals, to a user based at least in part on said abstract.

2. The method of claim 1, further comprising excluding redundant phrases during said generation of an abstract.

3. The method of claim 1, wherein generating said abstract is performed independent from a given query.

4. The method of claim 1, wherein generating said abstract further comprises selecting no more than a single phrase per sentence based at least in part on said utility function.

5. The method of claim 1, wherein said abstract comprises a subset of text from said electronic document.

6. The method of claim 1, wherein said utility metric comprises at least one of a measure of readability of a given phrase, a measure of relevance of a given phrase with respect to a given query, a measure of relevance of a given phrase with respect to said electronic document, and/or a measure of length of a given phrase.

7. The method of claim 1, wherein generating said abstract further comprises processing said utility metric based at least in part on a Viterbi-like algorithm to select a sequence of phrases.

8. The method of claim 1, further comprising:

extracting said set of two or more sentences from said electronic document;

determining a relevance of said two or more sentences associated with respect to said electronic document; and

ranking said two or more sentences into said ranked order based at least in part on said determined relevance.

9. The method of claim 1, further comprising:

extracting said set of two or more sentences from said electronic document;

determining an intent of a given query;

determining a query-dependent relevance of said two or more sentences with respect to a given query;

determining a degree to which said two or more sentences summarize said electronic document;

determining a relevance of said two or more sentences based at least in part upon said intent of a given query, said query-dependent relevance, and said degree to which said two or more sentences summarize said electronic; and

ranking said two or more sentences into said ranked order based at least in part on said determined relevance.

10. The method of claim 1, further comprising:

generating said search result in response to a given query, wherein said search result comprises at least one search result summary, and wherein said search result summary comprises at least one type of information portion selected from a group of distinguishable information portions comprising a title, an abstract, a link, and an object.

11. An article comprising:

a storage medium comprising machine-readable instructions stored thereon which, if executed direct a computing platform to;

extract a group of two or more sets of phrases, represented within a computing platform by digital electronic signals, from a set of two or more sentences associated with an electronic document, wherein a single sentence of said two or more sentences is associated with a single set of one or more phrases; and

generate an abstract, represented within a computing platform by digital electronic signals, of said electronic document using at least one phrase from said group of two or more sets of phrases based at least in part on a utility metric associated with said group of two or more sets of phrases as well as on a ranked order of said two or more sentences; and

transmit a search result, represented within a computing platform by digital electronic signals, to a user based at least in part on said abstract.

12. The article of claim 11, wherein said machine-readable instructions, if executed by a computing platform, further direct a computing platform to exclude redundant phrases during said generation of an abstract.

13. The article of claim 11, wherein said generation of an abstract is performed independent from a given query.

14. The article of claim 11, wherein said generation of an abstract further comprises selection of no more than a single phrase per sentence based at least in part on said utility function.

15. The article of claim 11, wherein said utility metric comprises at least one of a measure of readability of a given phrase, a measure of relevance of a given phrase with respect to a given query, a measure of relevance of a given phrase with respect to said electronic document, and/or a measure of length of a given phrase.

16. The article of claim 11, wherein said machine-readable instructions, if executed by a computing platform, further direct a computing platform to:

extract said set of two or more sentences from said electronic document;

determine an intent of a given query;

determine a query-dependent relevance of said two or more sentences with respect to a given query;

determine a degree to which said two or more sentences summarize said electronic document;

determine a relevance of said two or more sentences based at least in part upon said intent of a given query, said query-dependent relevance, and said degree to which said two or more sentences summarize said electronic;

rank said two or more sentences into said ranked order based at least in part on said determined relevance;

process said utility metric based at least in part on a Viterbi-like algorithm to select a sequence of phrases during said generation of an abstract;

wherein said abstract comprises a subset of text from said electronic document; and

generate said search result in response to a given query, wherein said search result comprises at least one search result summary, and wherein said search result summary comprises at least one type of information portion selected from a group of distinguishable information portions comprising a title, an abstract, a link, and an object.

17. An apparatus comprising:

a computing platform, said computing platform being adapted to;

extract a group of two or more sets of phrases, represented within a computing platform by digital electronic signals, from a set of two or more sentences associated with an electronic document, wherein a single sentence of said two or more sentences is associated with a single set of one or more phrases;

generate an abstract, represented within a computing platform by digital electronic signals, of said electronic document using at least one phrase from said group of two or more sets of phrases based at least in part on a utility metric associated with said group of two or more sets of phrases as well as on a ranked order of said two or more sentences; and

transmit a search result, represented within a computing platform by digital electronic signals, to a user based at least in part on said abstract.

18. The apparatus of claim 17, wherein said machine-readable instructions, if executed by a computing platform, further direct a computing platform to exclude redundant phrases during said generation of an abstract.

19. The apparatus of claim 17, wherein said generation of an abstract is performed independent from a given query.

20. The apparatus of claim 17, wherein said utility metric comprises at least one of a measure of readability of a given phrase, a measure of relevance of a given phrase with respect to a given query, a measure of relevance of a given phrase with respect to said electronic document, and/or a measure of length of a given phrase.