Automatic diagnosis of search relevance failures

• US 8,041,710 B2
• Filed: 11/13/2008
• Issued: 10/18/2011
• Est. Priority Date: 11/13/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Yin, Xiaoxin, Fan, Jinliang, Tan, Wenzhao, Wang, Yi-Min, He, Li-wei
• Primary Examiner(s): Le; Debbie
• Application Number:
  US12/270,201
• Publication Number:
  US 20100121841A1
• Time in Patent Office: 1,069 Days
• Field of Search: 707/707, 707/709, 707/711, 707/722, 707/748, 707/999.003, 707/999.004, 707/999.006
• US Class Current: 707/722
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

Search relevance failures are diagnosed automatically. Users presented with unsatisfactory search results can report their dissatisfaction through various mechanisms. Dissatisfaction reports can trigger automatic investigation into the root cause of such dissatisfaction. Based on the identified root cause, a search engine can be modified to resolve the issue creating dissatisfaction thereby improving search engine quality.

Claims

1. A system for diagnosis of search relevance failures, the system comprising:

memory;

at least one processor communicatively coupled to the memory;

a dissatisfaction report generator component to automatically generate search result dissatisfaction reports, the dissatisfaction report generator component including;

a human judgment component stored in the memory and executed by the at least one processor to;

determine a difference in relevance of search results provided by multiple search engines in response to an associated query; and

generate one or more search result dissatisfaction reports based in part on the difference; and

a click log component stored in the memory and executed by the at least one processor to;

analyze recorded clicks on search results in one or more query click logs;

determine if one or more portions of the search results are irrelevant to the associated query based in part on the analyzing of the recorded clicks; and

generate one or more search result dissatisfaction reports if one or more portions of the search results are determined to be irrelevant;

a collection component that acquires the search result dissatisfaction reports; and

a diagnostic component that automatically identifies one or more potential causes of reported dissatisfaction to facilitate resolution thereof.

2. The system of claim 1, the diagnostic component employs a decision diagram to automate identification of the causes.

3. The system of claim 1, the diagnostic component employs black-box techniques to identify the one or more potential causes.

4. The system of claim 1, further comprising a component that classifies in accordance with a number of predetermined classes or categories.

5. The system of claim 4, the categories address one of crawler, index, document candidate set, ranking, or diversity constraint issues.

6. The system of claim 1, the dissatisfaction reports identify relevant URLs that are missing from search results.

7. The system of claim 1, the dissatisfaction reports identify duplicate URLs in search results.

8. The system of claim 1, the collection component acquires automatically generated dissatisfaction reports.

9. The system of claim 1, further comprising a data store that houses the dissatisfaction reports to serve as test suites.

10. The system of claim 1, further comprising a correction component that automatically augments a search engine in accordance with the one or more identified causes to address the dissatisfaction report.

11. A search engine diagnostic method implemented by a computer, the method comprising:

receiving a dissatisfaction report identifying a missing uniform resource locator (URL) in search results in response to a query, the missing URL being relevant to the query; and

diagnosing root cause of the missing URL automatically by the computer by performing actions including;

determining whether the missing URL is in a candidate set for the query by determining keywords of the query found at a document referenced by the missing URL;

determining that one or more portions of a document referenced by the missing URL are not properly indexed;

determining that at least one of the one or more portions of the document are relevant to the query;

categorizing the root cause; and

initiating search engine modification to address the missing URL as a function of the categorized cause.

12. The method of claim 11, further comprising identifying a class of specified causes to which the root cause belongs, and forwarding the dissatisfaction report and the root cause to an entity associated with the identified class for resolution.

13. The method of claim 11, the determining keywords of the query being based in part on at least one of issuing the query at the missing URL or issuing one or more queries containing one or more of the keywords of the query at the missing URL.

14. The method of claim 11, the actions further including determining whether an alteration of the query results in presentation of the missing URL in corresponding search results, the alteration of the query including at least one of:

acronym expansion/contraction;

spell corrections;

stemming;

word breaking;

or

equivalence substitution.

15. The method of claim 11, the actions further including determining whether a core ranking of the search results was computed incorrectly causing the missing URL to be ranked below a ranking threshold or causing the missing URL to be absent in the core ranking.

16. The method of claim 11, the actions further comprising determining whether the missing URL is being excluded as a result of a host-based diversity constraint.

17. A method implemented by a computer for improving search engine quality, the method comprising:

receiving a dissatisfaction report that identifies a missing URL for a search query, the missing URL being relevant to the search query;

diagnosing automatically by the computer a root cause that prevented presentation of the missing URL in search results by performing actions including;

determining whether the missing URL is in a candidate set for the query based in part on determining keywords of the query found at a document referenced by the missing URL;

determining whether an alteration of the query results in presentation of the missing URL in corresponding search results; and

determining whether a portion of the document referenced by the missing URL is;

improperly indexed; and

relevant to the query;

categorizing the root cause; and

initiating search engine modification to address the missing URL as a function of the categorized cause.

18. The method of claim 17, further comprising extracting the dissatisfaction report from an unstructured format.

19. The method of claim 17, comprising notifying a person or team associated with the categorized root cause of the dissatisfaction report.