Search result inputs using variant generalized queries

• US 9,110,975 B1
• Filed: 11/02/2006
• Issued: 08/18/2015
• Est. Priority Date: 11/02/2006
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Diligenti, Michelangelo, Kim, Hyung-Jin, Stets, Robert J. Jr.
• Primary Examiner(s): Badawi, Sherief
• Assistant Examiner(s): RAHMAN, SABANA
• Application Number:
  US11/556,100
• Time in Patent Office: 3,211 Days
• Field of Search: 707/3, 707/104.1, 707/769, 707/723, 707/748, 704/1, 705/14
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3325 Reformulation based on results of preceding query
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

Systems, methods and computer program products for generalizing a user-submitted query by forming one or more variants of the user-submitted query to generate one or more other queries, each of the one or more other queries being different from the user-submitted query. A generalized quality of result statistic is derived for a first document from respective data associated with each of the other queries, each respective data being indicative of user behavior relative to the first document as a search result for the associated other query. The generalized quality of result statistic is provided as the quality of result statistic input to a document ranking process for the first document and the user-submitted query.

Claims

1. A computer-implemented method comprising:

receiving a user-submitted query;

generating an initial ranking of one or more documents responsive to the user-submitted query, the one or more documents including a first document responsive to the user-submitted query;

determining that the user-submitted query does not match any record in a model of queries and documents, the model comprising a plurality of records, each record specifying an association between a respective model query, a respective model document, and respective click data aggregated for all users that select the respective model document in response to the respective model query;

in response to determining that the user-submitted query does not match any record in the model, forming one or more variants of the user-submitted query to generate one or more other queries that each match a respective record in the model, each of the one or more other queries being different from the user-submitted query;

deriving a generalized quality of result statistic for the first document responsive to the user-submitted query from respective data associated in the model with one or more of the other queries, wherein;

the respective data for a given other query of the one or more of the other queries is associated with the matching respective record in the model and specifies aggregate user behavior for the first document when the first document was referenced in search results for the given other query, and

the user behavior in selection of a reference to the first document in the search results for the given other query; and

providing the generalized quality of result statistic as input to a subsequent document ranking process for the user-submitted query that generates, using the generalized quality of result statistic for the first document responsive to the user-submitted query, a subsequent ranking of the one or more documents responsive to the user-submitted query.

2. The method of claim 1 where:

the respective data associated in the model with one or more of the other queries comprises aggregated click data, wherein the aggregated click data is based on a weighted average of a count of long clicks for the first document in the search results for each of the one or more of the other queries.

3. The method of claim 1 where forming one or more variants of the user-submitted query includes:

attempting one or more variations of the user-submitting query until a match within the model has been located.

4. The method of claim 1 where the deriving includes:

modifying the quality of result statistic based on a level of certainty for each of the generated one or more queries.

5. The method of claim 1 where a variant of the user-submitted query is formed by performing one or more of:

replacing one or more terms in the user-submitted query with synonyms,

changing an order of one or more terms in the user-submitted query,

or replacing one or more query terms in user-submitted query with stems.

6. A computer program product tangibly embodied in a non-transitory a computer-readable storage device, the computer program product including instructions configured to cause one or more data processing apparatus to perform operations comprising:

receiving a user-submitted query;

generating an initial ranking of one or more documents responsive to the user-submitted query, the one or more documents including a first document responsive to the user-submitted query;

determining that the user-submitted query does not match any record in a model of queries and documents, the model comprising a plurality of records, each record specifying an association between a respective query, a respective document, and respective click data aggregated for all users that select the respective model document in response to the respective model query;

in response to determining that the user-submitted query does not match any query in the model, forming one or more variants of the user-submitted query to generate one or more other queries that each match a respective record in the model, each of the one or more other queries being different from the user-submitted query;

deriving a generalized quality of result statistic for the first document responsive to the user-submitted query from respective data associated in the model with one or more of the other queries, wherein;

the respective data for a given other query of the one or more of the other queries is associated with the matching respective record in the model and specifies aggregate user behavior for the first document when the first document was referenced in search results for the given other query, and

the user behavior in selection of a reference to the first document in the search results for the given other query; and

providing the generalized quality of result statistic as input to a subsequent document ranking process for the user-submitted query that generates, using the generalized quality of result statistic for the first document responsive to the user-submitted query, a subsequent ranking of the one or more documents responsive to the user-submitted query.

7. The program product of claim 6 where:

the respective data associated in the model with one or more of the other queries comprises aggregated click data, wherein the aggregated click data is based on a weighted average of a count of long clicks for the first document in the search results for each of the one or more of the other queries.

8. The program product of claim 6 where forming one or more variants of the user-submitted query includes:

attempting one or more variations of the user-submitting query until a match within the model has been located.

9. The program product of claim 6 where the deriving includes:

modifying the quality of result statistic based on a level of certainty for each of the generated one or more queries.

10. The program product of claim 6 where a variant of the user-submitted query is formed by performing one or more of:

replacing one or more terms in the user-submitted query with synonyms,

changing an order of one or more terms in the user-submitted query,

or replacing one or more query terms in user-submitted query with stems.

11. A system comprising:

a computer-readable storage device including instructions; and

one or more data processing apparatuses configured to execute the instructions and perform operations, comprising;

receiving a user-submitted query;

generating an initial ranking of one or more documents responsive to the user-submitted query, the one or more documents including a first document responsive to the user-submitted query;

determining that the user-submitted query does not match any record in a model of queries and documents, the model comprising a plurality of records, each record specifying an association between a respective query, a respective document, and respective click data aggregated for all users that select the respective model document in response to the respective model query;

in response to determining that the user-submitted query does not match any query in the model, forming one or more variants of the user-submitted query to generate one or more other queries that each match a respective record in the model, each of the one or more other queries being different from the user-submitted query;

deriving a generalized quality of result statistic for the first document responsive to the user-submitted query from respective data associated in the model with one or more of the other queries, wherein;

the respective data for a given other query of the one or more of the other queries is associated with the matching respective record in the model and specifies aggregate user behavior for the first document when the first document was referenced in search results for the given other query, and

the user behavior in selection of a reference to the first document in the search results for the given other query; and

providing the generalized quality of result statistic as input to a subsequent document ranking process for the user-submitted query that generates, using the generalized quality of result statistic for the first document responsive to the user-submitted query, a subsequent ranking of the one or more documents responsive to the user-submitted query.

12. The system of claim 11 where:

the respective data associated in the model with one or more of the other queries comprises aggregated click data, wherein the aggregated click data is based on a weighted average of a count of long clicks for the first document in the search results for each of the one or more of the other queries.

13. The system of claim 11 where forming one or more variants of the user-submitted query includes:

attempting one or more variations of the user-submitting query until a match within the model has been located.

14. The system of claim 11 where the deriving includes:

modifying the quality of result statistic based on a level of certainty for each of the generated one or more queries.

15. The system of claim 11 where a variant of the user-submitted query is formed by performing one or more of:

replacing one or more terms in the user-submitted query with synonyms,

changing an order of one or more terms in the user-submitted query,

or replacing one or more query terms in user-submitted query with stems.