RE-RANKING TOP SEARCH RESULTS

• US 20100262612A1
• Filed: 04/09/2009
• Published: 10/14/2010
• Est. Priority Date: 04/09/2009
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Burges, Christopher John Champness, El-Saban, Motaz Ahmad, Wu, Qiang
• Granted Patent: US 8,661,030 B2
• Time in Patent Office: Days
• US Class Current: 707/752
• CPC Class Codes: G06F 16/338 Presentation of query results
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

The claimed subject matter provides a system and/or a method that facilitates generating sorted search results for a query. An interface component can receive a query in a first language. A first ranker can be trained from a portion of data related to a second language. A second ranker can correspond to the first language, wherein the second ranker is untrained due to a limited amount of data related to the first language. A sorting component can invoke the first ranker to generate and order a pre-defined number of search results for the received query and subsequently invoke the second ranker to the pre-defined number of search results to generate a re-ordered number of search results in the first language for the received query.

Claims

1. A system that facilitates generating sorted search results for a query, comprising:

an interface component that receives a query in a first language;

a first ranker that is trained from a portion of data related to a second language;

a second ranker that corresponds to the first language, the second ranker is not sufficiently trained due to a limited amount of data related to the first language; and

a sorting component that invokes the first ranker to generate and order a pre-defined number of search results for the received query and subsequently invokes the second ranker to the pre-defined number of search results to generate a re-ordered number of search results in the first language for the received query.

2. The system of claim 1, further comprising a search engine that incorporates at least one of the sorting component, the first ranker, or the second ranker.

3. The system of claim 1, the sorting component employs a cascading ranking structure that includes the first ranker and the second ranker, wherein the output of the first ranker is an input for the second ranker in order to provide the re-ordered number of search results.

4. The system of claim 1, the sorting component leverages a hybrid training set of data related to the first ranker and the second ranker in order to provide the re-ordered number of search results.

5. The system of claim 1, the first ranker is trained by a portion of data that is greater than the portion of data utilized to train the second ranker and the first ranker is trained using enough data in comparison to the second ranker.

6. The system of claim 1, the second ranker is less trained in comparison to the first ranker, wherein the first ranker is trained and experimentally proven to provide a sufficiently relevant search result for a query to surface an initial candidate list.

7. The system of claim 1, the second ranker does not include sufficient training data in order to experimentally prove that the second ranker provided search results are relevant to a query.

8. The system of claim 1, further comprising a query evaluator that evaluates the query in real time to identify the first language utilizing an originating Internet Protocol (IP) address.

9. The system of claim 1, the query received is in at least one of a written form or a spoken form.

10. The system of claim 1, further comprising a ranker selector that selects at least one of the first ranker or the second ranker to invoke based upon the language of the query.

11. The system of claim 10, the ranker selector invokes a cascading technique with the first ranker and the second ranker based upon an amount of training data for the second ranker.

12. The system of claim 1, the sorting component employs a re-ordering with the first ranker and the second ranker on a client side for privacy.

13. The system of claim 1, at least one of the first ranker or the second ranker utilizes a feature vector to generate a relevancy score to provide at least one search result.

14. The system of claim 1, the query relates to a data search associated with at least one of the Internet, a data store, a database, a network, a website, or a server.

15. The system of claim 1, further comprising the first language is at least one of a spoken language, a written language, a typed language, a market, or a market segment.

16. The system of claim 15, further comprising the second language is disparate from the first language and is at least one of a spoken language, a written language, a typed language, a market, or a market segment.

17. A computer-implemented method that facilitates generating a search result for a received query, comprising:

receiving a query in a first language;

employing a language ranker corresponding to a second language to generate a set of search results for the received query; and

employing a language ranker corresponding to the first language to generate a re-ordering of the set of search results.

18. The method of claim 17, further comprising:

evaluating a ranker corresponding to the first language to ascertain if the ranker is sufficiently trained;

selecting a ranker corresponding to the second language based upon the evaluation; and

utilizing at least two rankers to provide a cascading ranking structure to produce the re-ordering of the set of search results in the event the ranker of the query language is insufficiently trained to provide accurate search results.

19. The method of claim 17, further comprising the language ranker corresponding to the second language is trained by a set of data that is greater than a set of training data utilized to train the language ranker corresponding to the first language.

20. A computer-implemented system that facilitates generating relevant search results for a query, comprising:

means for receiving a query in a first language;

means for evaluating a second ranker associated with the first language to ascertain if the second ranker is experimentally proven to generate relevant and ordered search results;

means for selecting a first ranker based upon the evaluation of the second ranker, wherein the first ranker relates to a second language and is trained with a first amount of data that is greater than a second amount of data related to training the second ranker;

means for utilizing the first ranker to generate a first set of ordered search results for the query in the first language;

means for invoking the second ranker on the first set of ordered search results to generate a second set of re-ordered search results for the query in the first language, wherein the second ranker is invoked on a pre-defined number of the first set of ordered search results;

means for directly presenting the second set of re-ordered search results to an entity that initiated the query; and

means for utilizing the first ranker and the second ranker with a search engine to satisfy the query.