Sentiment detection as a ranking signal for reviewable entities

• US 8,417,713 B1
• Filed: 12/05/2007
• Issued: 04/09/2013
• Est. Priority Date: 12/05/2007
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Blair-Goldensohn, Sasha, Hannan, Kerry, McDonald, Ryan, Neylon, Tyler, Reynar, Jeffrey C.
• Primary Examiner(s): Mahmoudi, Tony
• Assistant Examiner(s): Chau, Dung K
• Application Number:
  US11/951,155
• Time in Patent Office: 1,952 Days
• Field of Search: None
• US Class Current: 707/751
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/248 Presentation of query results
• CPC Class Codes: G06F 16/335 Filtering based on additional data, e.g. user or group profiles filtering in web context G06F16/9535, G06F16/9536
• CPC Class Codes: G06F 16/90335 Query processing
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 16/9538 Presentation of query results

Abstract

A method, a system and a computer program product for ranking reviewable entities based on sentiment expressed about the entities. A plurality of review texts are identified wherein each review text references an entity. A plurality of sentiment scores associated with the plurality of review texts are generated, wherein each sentiment score for a review text indicates a sentiment directed to the entity referenced by the review text. A plurality of ranking scores for the plurality of entities are generated wherein each ranking score is based at least in part on one or more sentiment scores associated with one or more review texts referencing the entity. A plurality of search results associated with the plurality of entities are displayed based at least in part on the ranking scores.

Claims

1. A computer-implemented method for ranking reviewable entities comprising:

using at least one processor and memory to perform steps comprising;

identifying a plurality of review texts, wherein each review text references at least one entity from a plurality of entities;

generating a plurality of sentiment scores based on the plurality of review texts, wherein each sentiment score for a review text indicates a sentiment directed to an entity referenced by the review text;

identifying a plurality of reviews, the reviews comprising ratings of the plurality of entities, the ratings separate from the review texts;

determining, from the plurality of reviews, ratings associated with the plurality of entities;

determining values indicating likelihoods that the review texts reference a particular one of the plurality of entities;

generating ranking scores for corresponding ones of the entities, wherein the ranking score of an entity is based upon the sentiment scores associated with review texts referencing the entity, the values indicating likelihoods that the review texts reference the entity, and the ratings associated with the entity in the plurality of reviews;

ranking the entities according to their associated ranking scores; and

storing the plurality of ranking scores.

2. The method of claim 1, further comprising displaying a plurality of search results associated with the plurality of entities based at least in part on the ranking scores.

3. The method of claim 2, further comprising monitoring a plurality of user interactions with the search results and generating ranking scores for the plurality of entities based at least in part on the plurality of user interactions with the search results.

4. The method of claim 3, wherein the generating ranking scores for corresponding ones of the entities comprises generating the ranking scores based at least in part on the plurality of user interactions with the search results comprises generating a mixture model based on the plurality of sentiment scores, a plurality of ratings from a plurality of structured reviews referencing the plurality of entities and the plurality of user interactions.

5. The method of claim 1, wherein the generating the ranking scores for corresponding ones of the entities comprises generating the ranking scores based on a first weight associated with at least a first sentiment score associated with a review text referencing the entity and a second weight associated with at least a first rating from a review referencing the entity.

6. The method of claim 1, further comprising:

determining a value that indicates a likelihood that a review text includes a sentiment directed to one of the plurality of entities; and

wherein

the ranking score for the one of the plurality of entities is further based on the value that indicates the likelihood that the review text includes the sentiment directed to one of the plurality of entities.

7. A non-transitory computer-readable storage medium encoded with computer program code for ranking reviewable entities, the computer program code comprising:

computer program code for identifying a plurality of review texts, wherein each review text references at least one entity from a plurality of entities;

computer program code for generating a plurality of sentiment scores based on the plurality of review texts, wherein each sentiment score for a review text indicates a sentiment directed to an entity referenced by the review text;

computer program code for identifying a plurality of reviews, the reviews comprising ratings of the plurality of entities, the ratings separate from the review texts;

computer program code for determining, from the plurality of reviews, ratings associated with the plurality of entities;

computer program code for determining values indicating likelihoods that the review texts reference a particular one of the plurality of entities;

computer program code for generating ranking scores for corresponding ones of the entities, wherein the ranking score of an entity is based upon the sentiment scores associated with review texts referencing the entity, the values indicating likelihoods that the review texts reference the entity, and the ratings associated with the entity in the plurality of reviews;

computer program code for ranking the entities according to their associated ranking scores; and

computer program code for storing the plurality of ranking scores.

8. The storage medium of claim 7, further comprising computer program code for displaying a plurality of search results associated with the plurality of entities based at least in part on the ranking scores.

9. The storage medium of claim 8, further comprising computer program code for monitoring a plurality of user interactions with the search results and computer program code for generating ranking scores for the plurality of entities based at least in part on the plurality of user interactions with the search results.

10. The storage medium of claim 9, wherein the computer program code for generating ranking scores for corresponding ones of the entities comprises computer program code for generating the ranking scores based at least in part on the plurality of user interactions with the search results comprises computer program code for generating a mixture model based on the plurality of sentiment scores, a plurality of ratings from a plurality of structured reviews referencing the plurality of entities and the plurality of user interactions.

11. The storage medium of claim 7, wherein the computer program code for generating the ranking scores for corresponding ones of the entities comprises computer program code for generating the ranking scores based on a first weight associated with at least a first sentiment score associated with a review text referencing the entity and a second weight associated with at least a first rating from a structured review referencing the entity.

12. The storage medium of claim 7, further comprising computer program code for:

determining a value that indicates a likelihood that a review text includes a sentiment directed to one of the plurality of entities; and

wherein

the ranking score for the one of the plurality of entities is further based on the value that indicates the likelihood that the review text includes the sentiment directed to one of the plurality of entities.

13. A system for ranking reviewable entities, the system comprising:

at least one processor for executing instructions in program modules, the program modules including;

a text selection module to identify a plurality of review texts, wherein each review text references at least one entity from a plurality of entities;

a sentiment score module to generate a plurality of sentiment scores based on the plurality of review texts, wherein each sentiment score for a review text indicates a sentiment directed to an entity referenced by the review text;

a rating module to identify a plurality of reviews, the reviews comprising ratings of the plurality of entities, the ratings separate from the review texts, and to determine, from the plurality of reviews, ratings associated with the plurality of entities; and

a rank learning module to determine values indicating likelihoods that the review texts reference a particular one of the plurality of entities, and to generate ranking scores for corresponding ones of the entities, wherein the ranking score of an entity is based upon the sentiment scores associated with review texts referencing the entity, the values indicating likelihoods that the review texts reference the entity, and the ratings associated with the entity in the plurality of reviews, further to rank the entities according to their associated ranking scores, and to store the plurality of ranking scores in a ranking database.

14. The system of claim 13, further comprising a user interaction module to display a plurality of search results associated with the plurality of entities based at least in part on the ranking scores.

15. The system of claim 14, wherein the rank learning module generates the ranking scores for corresponding ones of the entities based on a first weight associated with at least a first sentiment score associated with a review text referencing the entity and a second weight associated with at least a first rating from a review referencing the entity.

16. The system of claim 14, wherein the user interaction module monitors a plurality of user interactions with the search results and the rank learning module generates ranking scores for the plurality of entities based at least in part on the plurality of user interactions with the search results.

17. The system of claim 16, wherein the rank learning module generates a mixture model based on the plurality of sentiment scores, a plurality of ratings from a plurality of structured reviews referencing the plurality of entities and the plurality of user interactions.

18. The system of claim 13, wherein the rank learning module

determines a value that indicates a likelihood that a review text includes a sentiment directed to one of the plurality of entities;

and wherein

the ranking score for the one of the plurality of entities is further based on the value that indicates the likelihood that the review text includes the sentiment directed to one of the plurality of entities.