Information retrieval

• US 7,599,928 B2
• Filed: 01/19/2006
• Issued: 10/06/2009
• Est. Priority Date: 07/23/2003
• Status: Expired due to Fees

• Current Assignee: University College Dublin
• Original Assignee: University College Dublin
• Inventors: Balfe, Evelyn, Smyth, Barry
• Primary Examiner(s): Corrielus; Jean M
• Assistant Examiner(s): Kerzhner; Aleksandr
• Application Number:
  US11/334,753
• Publication Number:
  US 20060173830A1
• Time in Patent Office: 1,356 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9538 Presentation of query results
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

A meta search engine receives a query from a user, and submits adapted queries to n search engines. The results are adapted and combined to provide a result set R′. In parallel, the meta search engine accesses a selected hit matrix populated with values for correspondence of queries q₁-q_m with web pages p_i-p_n. These values are the number of “hits” for pervious searches. The pages retrieved from a row for q_T are combined with the search engine results to provide a final result but which is ranked according to previous search experience. The hit matrix may be dynamically selected according to a user community. The query need not exactly match a hit matrix query as rows can be selected according to similarity to the current query q_T with weightings applied.

Claims

1. An information retrieval method carried out by a meta search engine communicating with n underlying search engines S₁ through S_n via n associated adapters, the method comprising the steps of:

the meta search engine receiving a query q_T from a user;

the adapters transforming the query q_T received by the meta search engine into queries q₁ through q_n in the formats of the associated underlying search engines S₁ through S_n and submitting the queries q₁ through q_n to the underlying search engines S₁ through S_n;

the underlying search engines S₁ through S_n performing searches for the corresponding queries q₁ through q_n to provide n corresponding result lists R₁ through R_n, and returning the result lists R₁ through R_n to the adapters, wherein each of the result lists R₁ through R_n consists of a plurality of results,

the adapters reformatting the result lists R₁ through R_n as result lists R′

₁ through R′

_n and returning the result lists R′

₁ through R′

_n to the meta search engine, wherein each of the result lists R′

₁ through R′

_n consists of a plurality of results;

the meta search engine merging the result lists R′

₁ through R′

_n to provide a result list R′

of ranked results;

the meta search engine performing relevancy processing in a manner which does not employ user-specific information, using at least one hit matrix H having in one dimension Web pages p historically selected by a community of users operating within a specific domain of interest and in another dimension query strings q from the same community of users as selected the Web pages p, wherein the value H_ij in each cell indicates the number of times a particular Web page p_j has been previously selected by a user in response to a particular query string q_i, to provide a second result list R″

of particular Web pages p_j selected in response to a particular query string q_i identical or related to the query q_T, the corresponding cells for which have a value H_ij that is non-zero;

the meta search engine performing post-processing of the result list R′

in a manner which does not employ user-specific information, the post-processing including the steps of the meta search engine;

merging the result lists R′ and

R″

to provide a merged result list R; and

providing the merged result list R to the user as a response to the query q_T.

2. The method as claimed in claim 1, wherein the relevancy processing is performed in parallel with searching by the underlying search engines.

3. The method as claimed in claim 1, wherein the relevancy processing comprises computing the hit number for each result p_j for the query q_i relative to a total of the hit numbers for the query q_i to provide a relevance score for each result p_j, and ranking the results according to said relevancy scores.

4. The method as claimed in claim 3, wherein the meta search engine combines the relevance score with another score.

5. The method as claimed in claim 3, wherein the meta search engine combines the relevance score with a sponsored listing score ranking score.

6. The method as claimed in claim 1, wherein the merging step comprises giving higher preference to results from the second result list R″

and avoiding duplicates.

7. The method as claimed in claim 1, wherein the meta search engine dynamically selects from a plurality of hit matrices.

8. The method as claimed in claim 1, wherein the meta search engine dynamically selects from a plurality of hit matrices, and performs the selection according to a community to which the user is allocated.

9. The method as claimed in claim 8, wherein the community to which the user is allocated is determined according to source of the query.

10. The method as claimed in claim 8, wherein a Web site from which the query originates is used to determine the community to which the user is allocated.

11. The method as claimed in claim 8, wherein the user results R are presented by the meta search engine together with information about the community associated with the selected hit matrix.

12. The method as claimed in claim 11, wherein the results are presented in usage-based clusters.

13. The method as claimed in claim 1, wherein the meta search engine selects a plurality of hit matrices and firstly presents to the user results arising from relevancy processing and merging with a primary hit matrix.

14. The method as claimed in claim 1, wherein the meta search engine applies weightings to the results p_j of the hit matrix according to similarity of the current query q_T with the associated query strings.

15. The method as claimed in claim 14, wherein the similarity is determined according to extent of query word overlap.

16. The method as claimed in claim 1, wherein the meta search engine performs case based reasoning in which each case comprises a query string and k associated historical results forming k pairs, each pair having an identifier and a relevance score.

17. The method as claimed in claim 16, wherein each case corresponds to a dimension of a hit matrix.

18. The method as claimed in claim 1,

wherein the meta search engine dynamically selects from a plurality of hit matrices H, and performs the selection according to a community to which the user is allocated;

wherein the community to which the user is allocated is determined according to source of the query; and

wherein a Web site from which the query originates is used to determine the community to which the user is allocated.

19. A computer program product, comprising a computer readable medium having a computer readable program code embodied therein, said computer readable program code adapted to implement the steps performed by the meta search engine in the method of claim 1 when executing on a digital processor.