Click-through log mining

• US 8,321,448 B2
• Filed: 10/10/2007
• Issued: 11/27/2012
• Est. Priority Date: 02/22/2007
• Status: Expired due to Fees

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Zeng, HuaJun, Hu, Jian, Li, Hua, Wang, Jian, Chen, Zeng
• Primary Examiner(s): Jalil, Neveen Abel
• Assistant Examiner(s): PEACH, POLINA G
• Application Number:
  US11/870,359
• Publication Number:
  US 20080208841A1
• Time in Patent Office: 1,875 Days
• Field of Search: None
• US Class Current: 707/767
• CPC Class Codes: G06F 16/3326 using relevance feedback from the user, e.g. relevance feedback on documents, documents sets, document terms or passages
• CPC Class Codes: G06F 16/3338 Query expansion
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

Click-through log mining is described. Raw search click-through log data is processed to generate ordered query keywords, utilizing an algorithm to expand user-submitted keywords to include high frequency user queries, managing the keywords for a keyword expansion file, analyzing the algorithm performance on a bidding criteria, and identifying related phrases with similar page-click behaviors for advertisements.

Claims

1. A method for providing keyword searches, implemented at least in part by a computing device, the method comprising:

processing raw search click-through log data based at least in part on set time intervals including a month, a day, and a year, to generate a click-through log;

generating ordered query keywords from the raw search click-through log data in which the ordered query keywords include query-uniform resource locator pairs;

constructing a bipartite graph based at least in part on the query-uniform resource locator pairs in which one set of vertices corresponds to queries and a second set of vertices corresponds to uniform resource locators;

based at least in part on the bipartite graph, utilizing an algorithm to compute

(a) similarities between queries in which the similarities of the queries are based at least in part on visiting similar web pages associated with the uniform resource locators, and

(b) similarities between the web pages associated with the uniform resource locators in which the similarities of the web pages are based at least in part on being visited by similar type queries to capture related phrases, wherein the algorithm comprises determining relationships between the queries and the web pages by iteratively computing similarities between the queries and the web pages;

identifying advertising keywords based at least in part on the similarities of the queries and the similarities of the web pages used to capture the related phrases by using a keyword expansion file associated with the set time intervals to expand the queries to include the advertising keywords according to a query-uniform resource locator (URL) correlation, the advertising keywords are based at least in part on a bidding criteria for advertisements; and

suggesting the related phrases that have similar page-click behaviors based at least in part on the bidding criteria for the advertisements.

2. The method of claim 1, wherein the similarities between the queries and the web pages are iteratively computed until a fixed point is reached.

3. The method of claim 1, further comprising using the similarities between the queries to look up the advertisement keywords within the click-through log to get a top number (N) most similar queries.

4. The method of claim 1, wherein:

the query-uniform resource locator (URL) correlation is learned from a query-URL mapping file; and

the algorithm further calculates the ordered generated keywords based at least in part on a correlated URL.

5. The method of claim 1, wherein the algorithm further comprises ranking related ordered generated keywords with a relevance score.

6. The method of claim 1, wherein the bidding criteria further comprises at least one of a click-through rate or a revenue per search.

7. The method of claim 1, further comprising sorting the ordered query keywords according to a query alphabet.

8. The method of claim 1, further comprising filtering the query-uniform resource locator (URL) word pairs that are noisy.

9. The method of claim 1, further comprising selecting at least one option to amend the keyword expansion file or to replace the keyword expansion file with new keywords.

10. A computer-readable storage media comprising computer-readable instructions executed on a computing device, the computer-readable instructions comprising instructions for:

processing raw search click-through log data based at least in part on set time intervals including a month, a day, and a year, to generate a click-through log;

generating ordered query keywords from the raw search click-through log data in which the ordered query keywords include query-uniform resource locator pairs;

constructing a bipartite graph based at least in part on the query-uniform resource locator pairs in which one set of vertices corresponds to queries and a second set of vertices corresponds to uniform resource locators;

based at least in part on the bipartite graph, utilizing an algorithm to compute

(a) similarities between queries in which the similarities of the queries are based at least in part on visiting similar web pages associated with the uniform resource locators, and

(b) similarities between the web pages associated with the uniform resource locators in which the similarities of the web pages are based at least in part on being visited by similar type queries to capture related phrases, wherein the algorithm comprises determining relationships between the queries and the web pages by iteratively computing similarities between the queries and the web pages;

identifying advertising keywords based at least in part on the similarities of the queries and the similarities of the web pages used to capture the related phrases for a keyword expansion file associated with the set time intervals to expand the queries to include the advertising keywords according to a query-uniform resource locator (URL) correlation, the advertising keywords are based at least in part on a bidding criteria for advertisements; and

suggesting the related phrases that have similar page-click behaviors based at least in part on the bidding criteria for the advertisements.

11. The computer-readable storage media of claim 10, wherein the similarities between the queries and the web pages are computed iteratively until a fixed point is reached.

12. The computer-readable storage media of claim 10, further comprising using the similarities between the queries to look up advertisement keywords within the click-through log to get a top number (N) most similar queries.

13. The computer-readable storage media of claim 10, wherein:

the query-uniform resource locator (URL) correlation is learned from a query-URL mapping file; and

the algorithm further calculates the ordered generated keywords based at least in part on a correlated URL.

14. A keyword suggestion system, comprising:

a processor;

a memory coupled to the processor, wherein the processor is configured for;

processing raw search click-through log data based at least in part on set time intervals including a month, a day, and a year, to generate a click-through log, the ordered query keywords include keywords that are frequently submitted by users;

generating ordered query keywords from the raw search click-through log data in which the ordered query keywords include query-uniform resource locator pairs;

constructing a bipartite graph based at least in part on the query-uniform resource locator pairs in which one set of vertices corresponds to queries and a second set of vertices corresponds to uniform resource locators;

based at least in part on the bipartite graph, utilizing an algorithm to compute

(a) similarities between queries in which the similarities of the queries are based at least in part on visiting similar web pages associated with the uniform resource locators, and

(b) similarities between the web pages associated with the uniform resource locators in which the similarities of the web pages are based at least in part on being visited by similar type queries to capture related phrases, wherein the algorithm comprises determining relationships between the queries and the web pages by iteratively computing similarities between the queries and the web pages;

identifying advertising keywords based at least in part on the similarities of the queries and the similarities of the web pages used to capture the related phrases by using a keyword expansion file associated with the set time intervals to expand the queries to include the advertising keywords according to a query-uniform resource locator (URL) correlation, the advertising keywords are based at least in part on a bidding criteria for advertisements, wherein the bidding criteria comprises at least one of a click-through rate or a revenue per search; and

suggesting the related phrases that have similar page-click behaviors based at least in part on the bidding criteria for the advertisements.

15. The system of claim 14, wherein the similarities between the queries and the web pages are computed iteratively until a fixed point is reached.

16. The system of claim 14, further comprising using the similarities between the queries to look up advertisement keywords within the click-through log to get a top number (N) most similar queries.

17. The system of claim 14, further comprising selecting at least one option to amend the keyword expansion file or to replace the keyword expansion file.