Systems and Methods to Facilitate Local Searches via Location Disambiguation

• US 20120117007A1
• Filed: 11/04/2010
• Published: 05/10/2012
• Est. Priority Date: 11/04/2010
• Status: Active Grant

• Current Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Shanahan, James G., Agrawal, Ritesh Jitendra
• Granted Patent: US 8,473,433 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06F 16/29 Geographical information databases
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 16/9537 Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Systems and methods use machine learning techniques to resolve location ambiguity in search queries. In one aspect, a dataset generator generates a training dataset using query logs of a search engine. A training engine applies a machine learning technique to the training dataset to generate a location disambiguation model. A location disambiguation engine uses the location disambiguation model to resolve location ambiguity in subsequent search queries.

Claims

1. A computer-implemented method, comprising:

receiving, in a computing device, a search query having location ambiguity;

identifying, by the computing device, an unambiguous location for the search query to resolve the location ambiguity using a location model generated using a machine learning technique;

determining, by the computing device, a search result corresponding to the search query directed to the unambiguous location; and

providing, by the computing device, the search result as a response to the search query having location ambiguity.

2. The method of claim 1, further comprising:

generating the location model from query logs using the machine learning technique.

3. The method of claim 2, wherein the machine learning technique comprises a decision tree learning technique.

4. The method of claim 2, wherein the machine learning technique comprises a gradient boosted decision tree learning technique.

5. The method of claim 2, wherein the generating of the location model comprises generating a training dataset from the query logs.

6. The method of claim 5, wherein the training dataset has at least one location feature, at least one personalization feature, and at least one search feature.

7. The method of claim 5, wherein the training dataset has training features comprising at least one selected from the group consisting of:

location features, personalization features, and search features.

8. The method of claim 7, wherein the location features are independent of queries recorded in the query logs.

9. The method of claim 8, wherein the location features comprise at least one of:

area, population, population density, demography, popularity, and business population.

10. The method of claim 8, wherein the personalization features are related to searchers of the queries recorded in the query logs.

11. The method of claim 10, wherein the personalization features comprise at least one of:

searcher location, searcher traveling direction, searcher traveling speed, searcher searched location, and location of searcher selected businesses.

12. The method of claim 8, wherein the search features are independent of searchers of the queries recorded in the query logs.

13. The method of claim 12, wherein the search features comprise at least one of:

topic-location co-occurrence frequency, time, day, and month.

14. The method of claim 5, wherein the generating of the training dataset comprises:

filtering the query logs to identify first queries without location ambiguity;

generating, from the first queries, second queries having location ambiguity;

identifying location candidates for the second queries; and

generating training targets based on comparing unambiguous locations specified in the first queries and corresponding location candidates for the second queries.

15. A computer readable medium storing instructions which, when executed on a computer, cause the computer to:

generate a training dataset from query logs of a search engine; and

apply a machine learning technique to the training dataset to generate a location model for location disambiguation.

16. The computer readable medium of claim 15, wherein the training dataset includes search hit values, business count values, and distances between searchers and location candidates.

17. A system, comprising:

a data storage device storing query logs recording search queries received in a search engine and response data indicating user interaction with search results responsive to the search queries;

a dataset generator coupled with the data storage device to generate a training dataset using the query logs and the response data;

a training engine communicatively coupled with the dataset generator to generate a location model using a machine learning technique; and

a location disambiguation engine coupled with the search engine to resolve location ambiguity in search requests using the location model.

18. The system of claim 17, further comprising:

a search engine coupled with the data storage device and configured to accept search queries, including search queries having location ambiguity, and to provide the search results.

19. The system of claim 17, wherein the dataset generator further comprises:

a filter to identify first queries in the query logs, the first queries having no location ambiguity, and the response data indicating that searchers have selected an item in each of search results responsive to the first queries;

an ambiguous query generator to generate second queries from the first queries via adding location ambiguity;

a location engine to determine location candidates for the second queries; and

a comparator to generate target values based on comparing unambiguous locations specified in the first queries and corresponding location candidates for the second queries.

20. The system of claim 17, wherein the training dataset includes at least one training feature independent of queries recorded in the query logs and at least one training feature independent of searchers.