QUERY SUGGESTION GENERATION

• US 20090187515A1
• Filed: 01/17/2008
• Published: 07/23/2009
• Est. Priority Date: 01/17/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Cucerzan, Silviu-Petru, Rounthwaite, Robert L., Chan, Joanna, Andrew, Galen, Park, Sooho, Buckley, Jamie Paul
• Granted Patent: US 7,984,004 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06F 16/3322 using system suggestions G06F16/3325 takes precedence
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

Described herein is a system that facilitates assigning indications of usefulness to query suggestions. The system includes a query suggestion generator component that receives a query and generates a query suggestion based at least in part upon the received query. A model component outputs an indication of usefulness with respect to the query suggestion, wherein the model component is a machine-learned model of user behavior with respect to query suggestions.

Claims

1. A system that facilitates outputting indications of usefulness of query suggestions, comprising the following computer-executable components:

a query suggestion generator component that receives a query and generates a query suggestion based at least in part upon the received query; and

a model component that outputs an indication of usefulness with respect to the query suggestion, wherein the model component is a machine-learned model of user behavior with respect to query suggestions.

2. The system of claim 1, wherein the indication of usefulness is a probability that the issuer of the query will select the query suggestion.

3. The system of claim 1, wherein the query suggestion generator component generates a plurality of query suggestions based at least in part upon the received query, and wherein the model component outputs an indication of usefulness for each of the plurality of query suggestions.

4. The system of claim 3, further comprising a display component that selects query suggestions to display to an issuer of the query based at least in part upon the indications of usefulness that correspond to the plurality of query suggestions.

5. The system of claim 1, wherein the model component comprises one or more of a logistic regression model or a multi-class logistic regression model.

6. The system of claim 1, further comprising:

a query receiver component that receives a partial query, wherein the partial query is received by the query suggestion generator component, the query suggestion component generates query suggestions based at least in part upon the received partial query.

7. The system of claim 1, further comprising:

an interface component that receives user input with respect to the received query and/or the query suggestion that is presented to an issuer of the query; and

a data collector component that collects data pertaining to the user input, wherein the collected data is used to refine the model component.

8. The system of claim of claim 1, further comprising:

a grouping component that groups query suggestions into a plurality of different groups, wherein the model component individually assigns an indication of usefulness to each of the groups of query suggestions.

9. The system of claim 8, wherein the groups are assembled based at least in part upon an amount of overlap between search results that correspond to query suggestions.

10. The system of claim 1, wherein the query suggestion generator component generates a plurality of query suggestions based at least in part upon the received query, and further comprising a reinforcement learning component that analyzes the model component to determine where additional data can be used to train the model component and performs a cost/benefit analysis with respect to candidate query suggestions, the reinforcement learning component selects a query suggestion to provide to an issuer of the query to accumulate additional data that is used to refine the model component.

11. The system of claim 1, wherein a mobile device comprises at least one of the query suggestion generator component and the model component.

12. The system of claim 1, further comprising a profiler component that receives data pertaining to the issuer of the query and assigns the issuer of the query to a profile, wherein parameters of the model component are weighted in accordance with the assigned profile.

13. The system of claim 1, wherein the model component analyzes features of at least one of the query or query suggestion when outputting the indication of usefulness, the features comprising at least one of:

a type of transformation of the query;

frequency of issuance of the query suggestion;

relative popularity between the query and the query suggestion;

frequency that the query suggestion follows the query in a search session;

or features pertaining to search results corresponding to at least one of the query or the query suggestion.

14. A method that facilitates determining an indication of usefulness of a query suggestion with respect to a query, comprising the following computer-executable acts:

receiving the query from a user;

generating the query suggestion based at least in part upon the received query; and

using a learned model of user behavior with respect to query suggestions to output an indication of usefulness of the query suggestion based at least in part upon the received query.

15. The method of claim 14, further comprising:

generating a plurality of query suggestions based at least in part upon the received query;

using the learned model of user behavior to individually output an indication of usefulness for each of the plurality of query suggestions; and

displaying query suggestions in an order that corresponds to indications of usefulness corresponding thereto.

16. The method of claim 14, wherein the indication of usefulness is a probability that the user will select the query suggestion.

17. The method of claim 14, further comprising:

generating a plurality of query suggestions based at least in part upon the received query;

creating multiple groups of query suggestions, wherein each of the multiple groups of query suggestions includes a plurality of query suggestions; and

using the learned model of user behavior to individually output an indication of usefulness for each of the multiple groups of query suggestions.

18. The method of claim 14, wherein the learned model of user behavior comprises one or more of a logistic regression model or a multi-class logistic regression model.

19. The method of claim 14, further comprising training the model of user behavior based at least in part upon user interaction with the results or other elements of the page returned for the query suggestion.

20. A computer-readable medium comprising instructions that, when executed, perform the following acts:

receiving user history data pertaining to query suggestions; and

using machine learning on the received user history data to model user behavior with respect to query suggestions.