Generating ranked search results using linear and nonlinear ranking models

• US 8,346,765 B2
• Filed: 06/14/2010
• Issued: 01/01/2013
• Est. Priority Date: 06/19/2009
• Status: Active Grant

• Current Assignee: Alibaba Group Holding Ltd.
• Original Assignee: Alibaba Group Holding Ltd.
• Inventors: Guo, Ning, Xing, Fei, Xie, Yuheng, Hou, Lei, Zhang, Qin
• Primary Examiner(s): Coby, Frantz
• Application Number:
  US12/802,816
• Publication Number:
  US 20100325105A1
• Time in Patent Office: 932 Days
• Field of Search: 707/723, 707/748, 707752-753, 707705-706, 707/722, 707724-725, 707/769
• US Class Current: 707/723
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/248 Presentation of query results
• CPC Class Codes: G06F 16/30 of unstructured textual data document management systems G06F16/93
• CPC Class Codes: G06F 16/438 Presentation of query results
• CPC Class Codes: G06F 16/44 Browsing; Visualisation therefor
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06F 16/957 Browsing optimisation, e.g. caching or content distillation

Abstract

Generating ranked search results includes receiving a plurality of matching information items that match a search request, ranking at least some of the plurality of matching information items using a linear ranking model that linearly combines a first plurality of feature values to obtain a first set of ranked results, ranking at least some of the first set of ranked results using a nonlinear ranking model that nonlinearly combines a second plurality of feature values to obtain a second set of ranked results, and provide a search response based on the second set of ranked results.

Claims

1. A method for generating ranked search results, comprising:

receiving a plurality of matching information items that match a search request;

ranking at least some of the plurality of matching information items using a linear ranking model that linearly combines a first plurality of feature values to obtain a first set of ranked results, wherein the linear ranking model combines the first plurality of feature values in a linear fashion using weight coefficients corresponding to the first plurality of feature values;

ranking at least some of the first set of ranked results using a nonlinear ranking model that nonlinearly combines a second plurality of feature values to obtain a second set of ranked results, wherein the nonlinear ranking model combines the second plurality of feature values in a nonlinear fashion using weight coefficients corresponding to the second plurality of feature values; and

providing a search response based on the second set of ranked results.

2. The method of claim 1, wherein the first set of ranked results includes a greater number of matching information items than the second set of ranked results.

3. The method of claim 1, wherein the plurality of matching information items are a second plurality of matching information items;

and the method further comprising;

receiving the search request;

generating a first plurality of matching information items;

preprocessing the first plurality of matching information items; and

selecting from the first plurality of matching information items the second plurality of matching information items.

4. The method of claim 3, wherein the first plurality of matching information items includes a greater number of matching information items than the second plurality of matching information items.

5. The method of claim 1, wherein the linear ranking model outputs a first ranking score and the nonlinear ranking model outputs a second ranking score.

6. The method of claim 1, wherein at least some of the first plurality of feature values used by the linear ranking model are the same as at least some of the second plurality of feature values used by the nonlinear ranking model.

7. The method of claim 1, wherein the first plurality of feature values are determined by the matching information, the search request, or both.

8. The method of claim 1, wherein the nonlinear ranking model is determined by training.

9. A system for generating ranked search results, comprising:

one or more processors configured to;

receive a plurality of matching information items that match a search request;

rank at least some of the plurality of matching information items using a linear ranking model that linearly combines a first plurality of feature values to obtain a first set of ranked results, wherein the linear ranking model combines the first plurality of feature values in a linear fashion using weight coefficients corresponding to the first plurality of feature values;

rank at least some of the first set of ranked results using a nonlinear ranking model that nonlinearly combines a second plurality of feature values to obtain a second set of ranked results, wherein the nonlinear ranking model combines the second plurality of feature values in a nonlinear fashion using weight coefficients corresponding to the second plurality of feature values; and

provide a search response based at least in part on the second set of ranked results;

one or more memories coupled to the one or more processors, configured to provide the processors with instructions.

10. The system of claim 9, wherein the first set of ranked results includes a greater number of matching information items than the second set of ranked results.

11. The system of claim 9, wherein the plurality of matching information items are a second plurality of matching information items;

and the one or more processors are further configured to;

receive the search request;

generate a first plurality of matching information items;

preprocess the first plurality of matching information items; and

select from the first plurality of matching information items the second plurality of matching information items.

12. The system of claim 11, wherein the first plurality of matching information items includes a greater number of matching information items than the second plurality of matching information items.

13. The system of claim 9, wherein the linear ranking model outputs a first ranking score and the nonlinear ranking model outputs a second ranking score.

14. The system of claim 9, wherein at least some of the first plurality of feature values used by the linear ranking model are the same as at least some of the second plurality of feature values used by the nonlinear ranking model.

15. The system of claim 9, wherein the first plurality of feature values are determined by the matching information, the search request, or both.

16. The system of claim 9, wherein the nonlinear ranking model is determined by training.