Modifying scoring data based on historical changes

• US 8,898,153 B1
• Filed: 09/14/2012
• Issued: 11/25/2014
• Est. Priority Date: 11/20/2009
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Kim, Hyung-Jin, Lopatenko, Andrei
• Primary Examiner(s): LU, KUEN S
• Application Number:
  US13/618,111
• Time in Patent Office: 802 Days
• Field of Search: 707/723, 707/803
• US Class Current: 707/723
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for modifying historical data. One method includes calculating time trend statistics for a document and a query during different time periods. The method further includes modifying a quality of result statistic for the document as a search result for the query by a factor based on the one or more time trend statistics. The method further includes providing the modified quality of result statistic as an input to a document ranking process. Another method includes calculating a difference score for statistics for a group of documents and a query for a first time period and a second time period. The method further includes modifying quality of result statistics for documents responsive to the query based on the difference score. The method further includes providing the modified statistics as an input to a document ranking process.

Claims

1. A computer-implemented method comprising:

calculating a plurality of quality of result statistics for a first document as a search result for a first query, each of the quality of result statistics corresponding to a different respective time period;

calculating a plurality of time trend statistics, the time trend statistics estimating changes in the quality of result statistics over time, wherein each of the time trend statistics comprises a quality of result difference between a respective first quality of result statistic and a respective second quality of result statistic of the quality of result statistics;

generating a first modified quality of result statistic by modifying a first quality of result statistic for the first document as a search result for the first query by a factor, where the factor is based on, at least, a rate of change of the time trend statistics; and

providing the first modified quality of result statistic as an input to a document ranking process for the first document and the first query.

2. The method of claim 1, further comprising verifying that the quality of result difference satisfies a statistically significant threshold before generating the first modified quality of result statistic.

3. The method of claim 2, wherein the factor is further based on how statistically significant the quality of result difference is.

4. The method of claim 1, further comprising:

generating a second modified quality of result statistic for the first document and a second different query according to the time trend statistics; and

providing the second modified quality of result statistic as an input to the document ranking process for the first document and the second query.

5. The method of claim 1, further comprising determining that a first past version of the first document during the first time period is different from a second past version of the first document during the second time period before modifying the first quality of result statistic.

6. The method of claim 5, wherein determining that the first past version of the first document is different from the second past version of the first document comprises calculating a similarity score that is a measure of similarities between shingles extracted from the first past version of the first document and shingles extracted from the second past version of the first document, and determining that the similarity score satisfies a threshold.

7. The method of claim 1, further comprising:

identifying a third query as related to the first document based on a quality of result statistic for the third query and the first document;

generating a third modified quality of result statistic for the third query and a second document by modifying a quality of result statistic for the third query and the second document by a factor, where the factor is based on the time trend statistics; and

providing the third modified quality of result statistic as an input to the document ranking process for the second document and the third query.

8. The method of claim 1, wherein each quality of result statistic is an estimate of a respective percentage of users that found the first document relevant to the first query out of a first total number of users who viewed the first document as a search result for the first query during a respective time period.

9. A system comprising:

one or more computers programmed to perform operations comprising;

calculating a plurality of quality of result statistics for a first document as a search result for a first query, each of the quality of result statistics corresponding to a different respective time period;

calculating a plurality of time trend statistics, the time trend statistics estimating changes in the quality of result statistics over time, wherein each of the time trend statistics comprises a quality of result difference between a respective first quality of result statistic and a respective second quality of result statistic of the quality of result statistics;

generating a first modified quality of result statistic by modifying a first quality of result statistic for the first document as a search result for the first query by a factor, where the factor is based on, at least, a rate of change of the time trend statistics; and

providing the first modified quality of result statistic as an input to a document ranking process for the first document and the first query.

10. The system of claim 9, wherein the operations further comprise verifying that the quality of result difference satisfies a statistically significant threshold before generating the first modified quality of result statistic.

11. The system of claim 10, wherein the factor is further based on how statistically significant the quality of result difference is.

12. The system of claim 9, wherein the operations further comprise:

generating a second modified quality of result statistic for the first document and a second different query according to the time trend statistics; and

providing the second modified quality of result statistic as an input to the document ranking process for the first document and the second query.

13. The system of claim 9, wherein the operations further comprise determining that a first past version of the first document during the first time period is different from a second past version of the first document during the second time period before modifying the first quality of result statistic.

14. The system of claim 13, wherein determining that the first past version of the first document is different from the second past version of the first document comprises calculating a similarity score that is a measure of similarities between shingles extracted from the first past version of the first document and shingles extracted from the second past version of the first document, and determining that the similarity score satisfies a threshold.

15. The system of claim 9, further comprising:

identifying a third query as related to the first document based on a quality of result statistic for the third query and the first document;

generating a third modified quality of result statistic for the third query and a second document by modifying a quality of result statistic for the third query and the second document by a factor, where the factor is based on the time trend statistics; and

providing the third modified quality of result statistic as an input to the document ranking process for the second document and the third query.

16. The system of claim 9, wherein each quality of result statistic is an estimate of a respective percentage of users that found the first document relevant to the first query out of a first total number of users who viewed the first document as a search result for the first query during a respective time period.

17. A computer-readable storage medium having instructions stored thereon that, when executed by one or more computers, cause the computers to perform operations comprising:

calculating a plurality of quality of result statistics for a first document as a search result for a first query, each of the quality of result statistics corresponding to a different respective time period;

calculating a plurality of time trend statistics, the time trend statistics estimating changes in the quality of result statistics over time, wherein each of the time trend statistics comprises a quality of result difference between a respective first quality of result statistic and a respective second quality of result statistic of the quality of result statistics;

generating a first modified quality of result statistic by modifying a first quality of result statistic for the first document as a search result for the first query by a factor, where the factor is based on, at least, a rate of change of the time trend statistics; and

providing the first modified quality of result statistic as an input to a document ranking process for the first document and the first query.

18. The storage medium of claim 17, wherein the operations further comprise verifying that the quality of result difference satisfies a statistically significant threshold before generating the first modified quality of result statistic.

19. The storage medium of claim 18, wherein the factor is further based on how statistically significant the quality of result difference is.

20. The storage medium of claim 17, wherein the operations further comprise:

generating a second modified quality of result statistic for the first document and a second different query according to the time trend statistics; and

providing the second modified quality of result statistic as an input to the document ranking process for the first document and the second query.

21. The storage medium of claim 17, wherein the operations further comprise determining that a first past version of the first document during the first time period is different from a second past version of the first document during the second time period before modifying the first quality of result statistic.

22. The storage medium of claim 21, wherein determining that the first past version of the first document is different from the second past version of the first document comprises calculating a similarity score that is a measure of similarities between shingles extracted from the first past version of the first document and shingles extracted from the second past version of the first document, and determining that the similarity score satisfies a threshold.

23. The storage medium of claim 17, further comprising:

identifying a third query as related to the first document based on a quality of result statistic for the third query and the first document;

generating a third modified quality of result statistic for the third query and a second document by modifying a quality of result statistic for the third query and the second document by a factor, where the factor is based on the time trend statistics; and

providing the third modified quality of result statistic as an input to the document ranking process for the second document and the third query.

24. The storage medium of claim 17, wherein each quality of result statistic is an estimate of a respective percentage of users that found the first document relevant to the first query out of a first total number of users who viewed the first document as a search result for the first query during a respective time period.