User Behavior Model for Contextual Personalized Recommendation

• US 20120295640A1
• Filed: 05/19/2011
• Published: 11/22/2012
• Est. Priority Date: 05/19/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Mei, Tao, Xu, Ying-Qing, Li, Shipeng, Zhang, Bo, Xu, Peng, Zhuang, Jinfeng
• Granted Patent: US 8,751,472 B2
• Time in Patent Office: Days
• US Class Current: 455/456.3
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/248 Presentation of query results
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

A user behavior model provides personalized recommendations based in part on time and location, particularly to users of mobile devices. Entity types are ranked according to relevance to the user. Example entity types are restaurant, hotel, etc. The relevance may be based on reference to a large-scale database containing queries from other users. Additionally, entities within each entity type may be ranked based on relevance to the user and the time and location context. A user interface may display a ranked list of entity types, such as restaurant, hotel, etc., wherein each entity type is represented by a highest-ranked entity with the entity type. Thus, the user interface may display a highest-ranked restaurant, a highest-ranked hotel, etc. Upon user selection of one such entity type the user interface is replaced with a second user interface, for example showing a ranked hierarchy of restaurants, headed by the highest-ranked restaurant.

Claims

1. A method, comprising:

receiving context information about a user of a mobile device;

ranking entity types according to relevance to the user, the relevance to the user of the ranked entity types based in part on mass user behavior defined within a query database and based in part on the context information;

ranking entities within each entity type according to relevance to the user, the relevance to the user of ranked entities within each entity type based in part on mass user behavior defined within the query database and based in part on the context information;

displaying, within a user interface, a listing of entity types according to the ranking of the entity types; and

representing each listed entity type with a most relevant entity within the entity type according to the ranking of entities within each entity type.

2. The method as recited in claim 1, wherein receiving context information comprises receiving a location of the mobile device of the user and a local time at the location.

3. The method as recited in claim 1, additionally comprising:

receiving input including selection of one of the listed entity types; and

replacing the user interface with a second user interface listing the ranked entities within the selected entity type.

4. The method as recited in claim 3, additionally comprising:

receiving input including selection of one of the listed ranked entities within the selected entity type.

5. The method as recited in claim 1, wherein ranking entity types and ranking entities within each entity type is based in part on similarity of the user of the mobile device to other users of other mobile devices, and is measured in part by:

a term frequency-inverse document frequency (tf-idf) weight to determine similarity between the user from whom input was received and the users who submitted searches stored in the query database;

or

a cosine similarity function to determine similarity between the user from whom input was received and the users who submitted searches stored in the query database.

6. The method as recited in claim 1, wherein the relevance to the user of ranked entity types and ranked entities within an entity type is based in part on weighting queries within the query database, the queries associated with entities within each entity type, wherein and the weighting is based in part on:

7. The method as recited in claim 1, wherein the relevance to the user of ranked entity types and the relevance to the user of ranked entities within an entity type are based in part on weighting queries within the query database, and wherein the weighting is based in part on summing a product having two terms, a first term being a function of a frequency of a query conducted by a user who has conducted the query and a second term being a function of a similarity between the user from whom input was received and the user who has conducted the query.

8. The method as recited in claim 1, wherein the relevance to the user of ranked entity types and the relevance to the user of ranked entities within an entity type are based in part on weighting queries within the query database, and wherein the weighting is based in part on:

identifying at least one user having conducted a query in the query database;

determining a frequency at which the user identified as having conducted the query in the query database has conducted the query in the query database;

determining a similarity between the user from whom the input was received and the user identified as having conducted the query in the query database; and

weighting each query in the query database based on the frequency and the similarity.

9. The method as recited in claim 1, wherein the relevance to the user of ranked entity types and the relevance to the user of ranked entities within an entity type are based in part on weighting queries within the query database, the queries associated with entities within each entity type, and wherein the weighting is based in part on:

a frequency at which a user who has conducted the query has executed the query; and

a similarity between the user who has conducted the query and the user from whom the input was received.

10. One or more computer-readable media storing computer-executable instructions that, when executed, cause one or more processors to perform acts comprising:

controlling the one or more processors configured with executable instructions to perform;

receiving context information about a user of a mobile device including a local time of the user and a location of the user;

ranking entity types according to relevance to the user, the relevance to the user of the ranked entity types based in part on mass user behavior defined within a query database and based in part on the context information;

ranking entities within each entity type according to relevance to the user, the relevance to the user of ranked entities within each entity type based in part on mass user behavior defined within the query database and based in part on the context information;

displaying, within a user interface, a listing of entity types according to the ranking of the entity types;

receiving input including selection of one of the listed entity types; and

replacing the user interface with a second user interface listing the ranked entities within the selected entity type.

11. The one or more computer-readable media as recited in claim 10, wherein the listing of entity types according to the ranking of the entity types is based in part on the local time of the user and the location of the user.

12. The one or more computer-readable media as recited in claim 10, wherein the query database comprises queries fitting circumstances of a plurality of different users of different mobile devices.

13. The one or more computer-readable media as recited in claim 10, additionally comprising:

representing each listed entity type with a most relevant entity within the entity type according to the ranking of entities within each entity type.

14. The one or more computer-readable media as recited in claim 13, additionally comprising:

receiving input including selection of one of the listed ranked entities within the selected entity type.

15. The one or more computer-readable media as recited in claim 10, wherein ranking entity types and ranking entities within each entity type is based in part on similarity of the user of the mobile device to other users of other mobile devices, and is measured in part by:

a term frequency-inverse document frequency (tf-idf) weight to determine similarity between the user from whom input was received and the users who submitted searches stored in the query database;

or

a cosine similarity function to determine similarity between the user from whom input was received and the users who submitted searches stored in the query database.

16. The one or more computer-readable media as recited in claim 10, wherein ranking entity types and ranking entities within each entity type is based in part on a weighting of queries in the query database, and is measured in part by:

17. The one or more computer-readable media as recited in claim 10, wherein each query in the query database is weighted, for use in ranking the entity types and ranking the entities within each entity type.

18. The one or more computer-readable media as recited in claim 10, wherein the relevance to the user of ranked entity types and ranked entities within an entity type is based in part on weighting queries within the query database, the queries associated with entities within each entity type, and wherein the weighting is based in part on:

a frequency at which a user who has conducted the query has executed the query; and

a similarity between the user who has conducted the query and the user from whom the input was received.

19. A method to model user behavior for contextual personalized recommendation, comprising:

controlling the one or more processors configured with executable instructions to perform;

receiving input, the input being based in part on a selection made from a list of options by a user of a mobile device and based in part on a current context, the list of options being displayed on a screen of the mobile device and being based in part on options available at a local time of the user and a location of the user;

searching a query database to obtain a query set based on the input, the query database including queries intended to fit circumstances of a plurality of different users of mobile devices, the searching resulting in a different query set depending on the location of the user and the local time of the user;

weighting queries in the query set, the weighting based in part on comparing the user of the mobile device to other users of other mobile devices, the comparing being based in part on;

identifying at least one user having conducted a query in the query set;

determining a frequency at which the user identified as having conducted the query in the query set has conducted the query in the query set;

determining a similarity between the user from whom the input was received and the user identified as having conducted the query in the query set; and

weighting each query in the query set based on the frequency and the similarity;

selecting a query from among the weighted queries in the query set;

searching a network based on the selected query; and

returning results based on the search of the network.

20. The method as recited in claim 19, wherein weighting queries in the query set comprises:

a term frequency-inverse document frequency (tf-idf) weight to determine similarity between the user from whom input was received and the users who submitted searches in the query set; and

a cosine similarity function to determine similarity between the user from whom input was received and users who submitted searches in the query set.