Phrase based snippet generation

• US 8,010,539 B2
• Filed: 01/25/2008
• Issued: 08/30/2011
• Est. Priority Date: 01/25/2008
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: McDonald, Ryan, Blair-Goldensohn, Sasha, Hannan, Kerry, Neylon, Tyler, Reynar, Jeffrey C.
• Primary Examiner(s): Corrielus; Jean M
• Application Number:
  US12/020,483
• Publication Number:
  US 20090193011A1
• Time in Patent Office: 1,313 Days
• Field of Search: 707/736, 707/741, 707/750, 707/999.003, 707/999.005, 704/9, 715/231, 705/10
• US Class Current: 707/750
• CPC Class Codes: G06F 16/345 Summarisation for human users
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

Disclosed herein is a method, a system and a computer product for generating a snippet for an entity, wherein each snippet comprises a plurality of sentiments about the entity. One or more textual reviews associated with the entity is selected. A plurality of sentiment phrases are identified based on the one or more textual reviews, wherein each sentiment phrase comprises a sentiment about the entity. One or more sentiment phrases from the plurality of sentiment phrases are selected to generate a snippet.

Claims

1. A computer-implemented method for generating a snippet for an entity, wherein the snippet comprises a plurality of sentiments about the entity, the method comprising:

selecting a plurality of textual reviews associated with the entity;

identifying, based on a domain-specific sentiment lexicon including sentiment phrases extracted from documents specific to a domain, a plurality of sentiment phrases from the plurality of textual reviews, wherein each identified sentiment phrase comprises a sentiment about the entity;

determining a plurality of frequency values for the identified plurality of sentiment phrases, wherein each frequency value represents a number of times a sentiment phrase is identified in the one or more of textual reviews;

selecting one or more sentiment phrases from the identified plurality of sentiment phrases based on the generated frequency values;

generating the snippet including the selected one or more sentiment phrases; and

storing the snippet.

2. The method of claim 1, further comprising displaying the snippet in association with a search result.

3. The method of claim 2, wherein the search result is displayed responsive to a search query comprising an entity type associated with the entity.

4. The method of claim 1, wherein identifying a plurality of sentiment phrases comprises identifying a noun phrase corresponding to a property of the entity and an adjective associated with the noun phrase.

5. The method of claim 4, wherein identifying a noun phrase corresponding to a property of the entity and an adjective associated with the noun phrase comprises identifying a match between a regular expression and a textual review.

6. The method of claim 4, further comprising:

generating a sentiment score for each of the plurality of sentiment phrases, wherein each sentiment score is based at least in part on the adjective associated with the noun phrase.

7. The method of claim 4, wherein selecting one or more sentiment phrases comprises:

determining at least a first group of sentiment scores based on the noun phrase; and

selecting a representative sentiment phrase based on the plurality of frequency values.

8. The method of claim 1, further comprising:

generating a sentiment score for each of the plurality of sentiment phrases; and

selecting one or more sentiment phrases from the plurality of sentiment phrases based at least in part on the sentiment scores.

9. The method of claim 1, wherein the sentiment phrases are comprised of a plurality of tokens and selecting one or more sentiment phrases comprises:

generating a redundancy metric, wherein each redundancy metric is based on a number of tokens shared between two sentiment phrases; and

selecting one or more sentiment phrases based on the redundancy metric.

10. The method of claim 1, further comprising:

associating a sentiment score with each of the identified plurality of sentiment phrases;

wherein the selection of the one or more sentiment phrases from the identified plurality of sentiment phrases is based on the corresponding sentiment scores.

11. The method of claim 1, further comprising

developing the domain-specific sentiment lexicon based on a domain-independent lexicon, the domain-independent sentiment lexicon including sentiment phrases extracted from domain independent documents.

12. The method of claim 11, wherein developing the domain-specific sentiment lexicon comprises:

determining sentiment scores for each of the documents that are specific to the domain;

selecting a subset of the domain-specific documents based on their sentiment scores;

selecting sentiment phrases from the subset of domain-specific documents; and

including in the domain-specific sentiment lexicon at least some of the sentiment phrases selected from the subset.

13. The method of claim 1, wherein the domain comprises a particular sphere of activity and the entity is of a type within the domain.

14. A computer readable storage medium comprising computer program code for generating a snippet for an entity, wherein the snippet comprises a plurality of sentiments about the entity, the computer program code comprising program code for:

selecting a plurality of textual reviews associated with the entity;

identifying, based on a domain-specific sentiment lexicon including sentiment phrases extracted from documents specific to a domain, a plurality of sentiment phrases from the plurality of textual reviews, wherein each identified sentiment phrase comprises a sentiment about the entity;

determining a plurality of frequency values for the identified plurality of sentiment phrases, wherein each frequency value represents a number of times a sentiment phrase is identified in the one or more of textual reviews;

selecting one or more sentiment phrases from the identified plurality of sentiment phrases based on the generated frequency values;

generating the snippet including the selected one or more sentiment phrases; and

storing the snippet.

15. The medium of claim 14, further comprising displaying the snippet in association with a search result.

16. The medium of claim 15, wherein the search result is displayed responsive to a search query comprising an entity type associated with the entity.

17. The medium of claim 14, wherein identifying a plurality of sentiment phrases comprises identifying a noun phrase corresponding to a property of the entity and an adjective associated with the noun phrase.

18. The medium of claim 17, wherein identifying a noun phrase corresponding to a property of the entity and an adjective associated with the noun phrase comprises identifying a match between a regular expression and a textual review.

19. The medium of claim 17, further comprising program code for:

generating a sentiment score for each of the plurality of sentiment phrases, wherein each sentiment score is based at least in part on the adjective associated with the noun phrase.

20. The medium of claim 17, wherein selecting one or more sentiment phrases comprises:

determining at least a first group of sentiment scores based on the noun phrase; and

selecting a representative sentiment phrase based on the plurality of frequency values.

21. The medium of claim 14, further comprising program code for:

generating a sentiment score for each of the plurality of sentiment phrases; and

selecting one or more sentiment phrases from the plurality of sentiment phrases based at least in part on the sentiment scores.

22. The medium of claim 14, wherein the sentiment phrases are comprised of a plurality of tokens and selecting one or more sentiment phrases comprises:

generating a redundancy metric, wherein each redundancy metric is based on a number of tokens shared between two sentiment phrases; and

selecting one or more sentiment phrases based on the redundancy metric.

23. A computer system for generating a snippet for an entity, wherein the snippet comprises a plurality of sentiments about the entity, the system comprising:

a database comprising a plurality of textual reviews associated with the entity;

a phrase extraction module adapted to

identify, based on a domain-specific sentiment lexicon including sentiment phrases extracted from documents specific to a domain, a plurality of sentiment phrases from the plurality of textual reviews, wherein each identified sentiment phrase comprises a sentiment about the entity; and

determine a plurality of frequency values for the identified plurality of sentiment phrases, wherein each frequency value represents a number of times a sentiment phrase is identified in the one or more of textual reviews; and

a snippet generation module adapted to;

select one or more sentiment phrases from the identified plurality of sentiment phrases based on the generated frequency values;

generate the snippet including the selected one or more sentiment phrases; and

store the snippet.

24. The system of claim 23, wherein the phrase extraction module is further adapted to identify a noun phrase corresponding to a property of the entity and an adjective associated with the noun phrase.

25. The system of claim 24, further comprising a sentiment score module adapted to:

generate a sentiment score for each of the plurality of sentiment phrases, wherein each sentiment score is based at least in part on the adjective associated with the noun phrase.

26. The system of claim 23, further comprising a sentiment score module adapted to generate a sentiment score for each of the plurality of sentiment phrases and wherein the phrase extraction module is adapted to select one or more sentiment phrases from the plurality of sentiment phrases based at least in part on the sentiment scores.