Framework for evaluating web search scoring functions

• US 8,060,497 B1
• Filed: 07/23/2009
• Issued: 11/15/2011
• Est. Priority Date: 07/23/2009
• Status: Expired due to Fees

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Wu, Yonghui, Zatsman, Misha, Haahr, Paul G., Cutts, Matthew D.
• Primary Examiner(s): Nguyen; Cam-Linh
• Application Number:
  US12/508,023
• Time in Patent Office: 845 Days
• Field of Search: 707/706, 707/722, 707/723, 707/735, 707/748, 707/752
• US Class Current: 707/723
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9538 Presentation of query results

Abstract

Methods, systems, and apparatus, including computer program products, for testing web search scoring functions. A query is received. A first and a second scoring function are selected by receiving search results responsive to the query; applying candidate scoring functions to the search results to determine scores for the search results for each candidate scoring function; identifying pairs of the candidate scoring functions, and calculating a diversity score for each of the pairs. A pair of candidate scoring functions is chosen from the one or more pairs of candidate scoring functions based on the diversity scores, and the alpha function is selected as the first scoring function and the beta function is selected as the second scoring function. The plurality of search results are presented in an order according to scores from the first scoring function and are presented in an order according to scores from the second scoring function.

Claims

1. A computer-implemented method comprising:

receiving a first query;

selecting, at a data processing apparatus, a first and a second candidate scoring function from a plurality of candidate scoring functions, the selecting comprising;

receiving a plurality of first search results responsive to the first query;

applying each of the candidate scoring functions to the plurality of first search results to determine respective scores for the first search results for the candidate scoring function;

identifying one or more pairs of candidate scoring functions from the candidate scoring functions, each pair comprising a respective alpha candidate scoring function and a respective beta candidate scoring function;

calculating a respective diversity score for each of the pairs of candidate scoring functions, where the diversity score measures diversity between a number of highest ranked first search results ordered according to the respective scores for the alpha candidate scoring function and a number of highest ranked first search results ordered according to the respective scores for the beta candidate scoring function; and

selecting the respective alpha candidate scoring function for a first pair of candidate scoring functions as the first scoring function and the respective beta candidate scoring function for the first pair as the second scoring function, where the first pair of candidate scoring functions is chosen from the one or more pairs of candidate scoring functions based on the respective diversity scores;

presenting the plurality of first search results ordered according to scores from the first scoring function; and

presenting the plurality of first search results ordered according to scores from the second scoring function.

2. The method of claim 1, further comprising:

receiving first input selecting the first scoring function over the second scoring function; and

generating an order for the plurality of scoring functions based on the first input.

3. The method of claim 2, further comprising:

receiving a second query;

receiving a plurality of second search results responsive to the second query;

selecting a third candidate scoring function and a fourth candidate scoring function from the plurality of scoring functions;

presenting the plurality of second search results ordered according to the third scoring function and presenting the plurality of second search results ordered according to the fourth scoring function;

receiving second input selecting the third scoring function over the fourth scoring function; and

wherein generating the order of the plurality of scoring functions is based on the first input and the second input.

4. The method of claim 3, further comprising detecting that the second input is suspicious, wherein:

generating the order of the plurality of scoring functions includes giving more weight to the first input than to the second input.

5. The method of claim 4, wherein detecting that an input is suspicious includes examining a history of inputs.

6. The method of claim 2, wherein the plurality of first search results have a first order according to the respective scores for the first candidate scoring function, the method further comprising:

receiving a notification that the first scoring function has been modified;

applying the modified first scoring function to the plurality of first search results to determine modified scores for the first search results, and generating a second order for the first search results according to the modified scores;

determining that the first order and the second order are different, and then modifying the order of the plurality of scoring functions so that the modified order of the plurality of scoring functions does not reflect the first input.

7. The method of claim 1, further comprising:

classifying the first query; and

using the classification to select the plurality of candidate scoring functions from a plurality of available scoring functions.

8. The method of claim 1, further comprising selecting the plurality of candidate scoring functions from a plurality of available scoring functions based on a respective score associated with each scoring function.

9. The method of claim 1, wherein calculating the diversity score for an alpha scoring function and a beta scoring function comprises comparing an alpha search result in a first position in an order of search results according to the scores for the alpha candidate scoring function and a beta search result in the first position in an order of search results according to the scores for the beta candidate scoring function.

10. The method of claim 1, wherein applying a candidate scoring function to the plurality of first search results further includes:

providing signals including information about the query and each search result in the plurality of first search results to the candidate scoring function.

11. A system comprising:

one or more computers comprising one or more processors and one or more memory devices;

the one or more computers programmed to perform operations comprising;

receiving a first query;

selecting a first and a second candidate scoring function from a plurality of candidate scoring functions, the selecting comprising;

receiving a plurality of first search results responsive to the first query;

applying each of the candidate scoring functions to the plurality of first search results to determine respective scores for the first search results for the candidate scoring function;

identifying one or more pairs of candidate scoring functions from the candidate scoring functions, each pair comprising a respective alpha candidate scoring function and a respective beta candidate scoring function;

calculating a respective diversity score for each of the pairs of candidate scoring functions, where the diversity score measures diversity between a number of highest ranked first search results ordered according to the respective scores for the alpha candidate scoring function and a number of highest ranked first search results ordered according to the respective scores for the beta candidate scoring function; and

selecting the respective alpha candidate scoring function for a first pair of candidate scoring functions as the first scoring function and the respective beta candidate scoring function for the first pair as the second scoring function, where the first pair of candidate scoring functions is chosen from the one or more pairs of candidate scoring functions based on the respective diversity scores;

presenting the plurality of first search results ordered according to scores from the first scoring function; and

presenting the plurality of first search results ordered according to scores from the second scoring function.

12. The system of claim 11, further programmed to perform operations comprising:

receiving first input selecting the first scoring function over the second scoring function; and

generating an order for the plurality of scoring functions based on the first input.

13. The system of claim 12, further programmed to perform operations comprising:

receiving a second query;

receiving a plurality of second search results responsive to the second query;

selecting a third candidate scoring function and a fourth candidate scoring function from the plurality of scoring functions;

presenting the plurality of second search results ordered according to the third scoring function and presenting the plurality of second search results ordered according to the fourth scoring function;

receiving second input selecting the third scoring function over the fourth scoring function; and

wherein generating the order of the plurality of scoring functions is based on the first input and the second input.

14. The system of claim 13, further programmed to perform operations comprising detecting that the second input is suspicious, wherein:

generating the order of the plurality of scoring functions includes giving more weight to the first input than to the second input.

15. The system of claim 14, wherein detecting that an input is suspicious includes examining a history of inputs.

16. The system of claim 12, wherein the plurality of first search results have a first order according to the respective scores for the first candidate scoring function, the system further programmed to perform operations comprising:

receiving a notification that the first scoring function has been modified;

applying the modified first scoring function to the plurality of first search results to determine modified scores for the first search results, and generating a second order for the first search results according to the modified scores;

determining that the first order and the second order are different, and then modifying the order of the plurality of scoring functions so that the modified order of the plurality of scoring functions does not reflect the first input.

17. The system of claim 11, further programmed to perform operations comprising:

classifying the first query; and

using the classification to select the plurality of candidate scoring functions from a plurality of available scoring functions.

18. The system of claim 11, further programmed to perform operations comprising selecting the plurality of candidate scoring functions from a plurality of available scoring functions based on a respective score associated with each scoring function.

19. The system of claim 11, wherein calculating the diversity score for an alpha scoring function and a beta scoring function comprises comparing an alpha search result in a first position in an order of search results according to the scores for the alpha candidate scoring function and a beta search result in the first position in an order of search results according to the scores for the beta candidate scoring function.

20. The system of claim 11, wherein applying a candidate scoring function to the plurality of first search results further includes:

providing signals including information about the query and each search result in the plurality of first search results to the candidate scoring function.