Page re-ranking system and re-ranking program to improve search result

• US 7,584,185 B2
• Filed: 01/12/2007
• Issued: 09/01/2009
• Est. Priority Date: 01/13/2006
• Status: Expired due to Fees

• Current Assignee: National Institute of Information and Communications Technology
• Original Assignee: National Institute of Information and Communications Technology
• Inventors: Tanaka, Katsumi, Jatowt, Adam, Kawai, Yukiko
• Primary Examiner(s): Corrielus; Jean M
• Application Number:
  US11/652,723
• Publication Number:
  US 20070174279A1
• Time in Patent Office: 963 Days
• Field of Search: 707/101, 707/10, 707/4, 707/3, 707/5, 707/6, 707/7, 707/102, 707/104.1
• US Class Current: 1/1
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

A page re-ranking system includes a super page producing part that produces a super page where page contents are combined between multiple versions for each of multiple Web pages that can be obtained as a search result page in compliance with a user'"'"'s query and to which a page ranking is created, a super page analyzing part that analyzes a covering degree of a topic representation that is contained in the super page produced by the super page producing part, and a re-ranking part that grants a renewed page ranking to each of the Web pages by comparing the analysis results obtained by the super page analyzing part between the super pages.

Claims

1. A page re-ranking system embedded in a memory and causing a processor to perform page re-ranking comprising:

a super page producing part that produces a super page wherein existing page contents are combined between multiple versions of the existing page from each of multiple Web pages that are obtained as a search result page in compliance with a user'"'"'s query on a topic and to which a page ranking is created;

a super page analyzing part that analyzes a covering degree of a topic representation that describes the query on the topic or is most related to the query on the topic and is contained in each of the super pages produced by the super page producing part; and

a re-ranking part that generates a renewed page ranking for each of the Web pages by comparing the analysis results obtained by the super page analyzing part between the super pages to provide a final ranking output.

2. The page re-ranking system described in claim 1, wherein

the covering degree of the topic representation contained in the super page of one of the multiple Web pages is a similarity degree or a difference degree concerning the topic representation between the super pages of the multiple Web pages.

3. The page re-ranking system of claim 2, wherein

the similarity degree between the super pages of the multiple Web pages is calculated by the following equation,

${\mathrm{HR}}_j=\sum _{}^{}$

i = 1 W ⁢

⁢

[ cos ⁡

( VS i , VS j ) * ( 1 + ɛ

* W - R i SE W ) * ( 1 + κ

* 1 Age i ) ]

where HR_j is a historical similarity degree of a Web page j, cos is a cosine similarity, VS_i is A vector of the super page of a Web page i, VS_j is a vector of the super page of the Web page j, R_i^SE is the search engine ranking of the Web page i, ε and

κ

are parameters, W is a total number of URLs as being an object to be re-ranked among a number of URLs of the search result obtained from the search engine, and Age_i is an age of the Web page i.

4. The page re-ranking system of claim 2 wherein the similarity degree between the super pages is calculated using a historical similarity of a first Web page, and a cosine similarity of a vector of a super page of the first Web page and a vector of a super page of a second Web page.

5. The page re-ranking system of claim 1, and

comprising a current page analyzing part that analyzes the covering degree of the topic representation that is contained in a current version of one of the multiple Web pages, wherein the re-ranking part generates a renewed page ranking for each of the multiple Web pages by comparing the analysis results obtained by the current page analyzing part between the current versions of the multiple Web pages.

6. The page re-ranking system of claim 5, wherein

the covering degree of the topic representation that is contained in the current version of one of the multiple Web pages is a similarity degree or a difference degree concerning the topic representation between the current versions of the multiple Web pages.

7. The page re-ranking system of claim 6, wherein

the similarity degree between the current versions of the multiple Web pages is calculated by the following equation,

$\begin{array}{c}{\mathrm{CR}}_j=\sum _{}^{}\end{array}$

i = 1 W ⁢

⁢

[ cos ⁡

( V i , V j ) * ( 1 + ɛ

* W - R i SE W ) ]

where CR_j is a current similarity degree of a Web page j, cos is a cosine similarity, V_i is a vector of a current version of the Web page i, V_j is a vector of a current version of the Web page j, R_i^SE is a search engine ranking of the Web page i, ε

is a parameter, W is a total number of URLs as being an object to be re-ranked among a number of URLs of the search result obtained from the search engine.

8. The page re-ranking system of claim 6 wherein the similarity degree between the current versions of the multiple Web pages is calculated using a current similarity degree of a first Web page, and a cosine similarity of a vector of a current version of the first Web page and a vector of a current version of a second Web page.

9. The page re-ranking system of claim 1, wherein

the super page producing part produces the super pages by referring to a Web archive part that memorizes multiple versions of Web pages that existed on the Internet in the past.

10. The page re-ranking system of claim 1, wherein

the search result page is obtained by a searching process by a Web search engine.

11. A page re-ranking program stored on a computer-readable storage medium, said page ranking program when executed by a computer causes a processor to perform page ranking comprising:

a super page producing part that produces a super page where page contents between multiple versions of each of multiple Web pages are combined for each of the multiple Web pages that are obtained as a search result page in compliance with a user'"'"'s query on a topic and to which a page ranking is provided;

a super page analyzing part that analyzes a covering degree of a topic representation that is contained in the super page produced by the super page producing part; and

a re-ranking part that creates a renewed page ranking for each of the multiple Web pages by comparing the analysis results between the super pages obtained by the super page analyzing part to provide a final ranking output of the search results to the user.

12. A search system for ranking relevant information in response to a user'"'"'s search of a topic on the Internet comprising;

a computer communicating with an Internet, a search engine, and a Web archive storing Web pages;

a user interface unit for entering a topic to be searched by the search engine on the Internet including the Web archive;

means for storing Web pages found by the search engine in response to the user'"'"'s topic;

a super page producing part that produces a super page wherein existing Web page contents between multiple versions of an existing Web page from each of multiple Web pages are obtained as a search result page in compliance with the user'"'"'s query on the topic and to which a page ranking is created for the individual Web pages;

a super page analyzing part analyzes the covering degree of a topic representation that describes the query on the topic or is most related to the query on the topic and is contained in each of the super pages produced by the super page producing part; and

a re-ranking part that creates a renewed page ranking for each of the multiple Web pages by comparing the analysis results obtained by the super page analyzing part between the super pages to provide a final ranking output to the user of the Web pages.

13. The search system described in claim 12, wherein

the covering degree of the topic representation contained in the super page of one of the multiple Web pages is a similarity degree or a difference degree concerning the topic representation between the super pages of the multiple Web pages.

14. The search system of claim 13, wherein

the similarity degree between the super pages of the multiple Web pages is calculated by the following equations

${\mathrm{HR}}_j=\sum _{}^{}$

i = 1 W ⁢

⁢

[ cos ⁡

( VS i , VS j ) * ( 1 + ɛ

* W - R i SE W ) * ( 1 + κ

* 1 Age i ) ]

where HR_j is a historical similarity degree of a Web page j, cos is a cosine similarity, VS_i is a vector of the super page of a Web page i, VS_j is a vector of the super page of the Web page j, R_i^SE is a search engine ranking of the Web page i, ε and

κ

are parameters, W is a total number of URLs as being an object to be re-ranked among a number of URLs of the search result obtained from the search engine, and Age_i is an age of the Web page i.

15. The search system of claim 13, and

comprising a current page analyzing part that analyzes a covering degree of the topic representation that is contained in a current version of one of the multiple Web pages, wherein the re-ranking part generates a renewed page ranking to each of the multiple Web pages by comparing the analysis results obtained by the current page analyzing part between the current versions of the multiple Web pages.

16. The search system of claim 15, wherein

the covering degree of the topic representation that is contained in the current version of one of the multiple Web pages is a similarity degree or a difference degree concerning the topic representation between the current versions of the multiple Web pages.

17. The search system of claim 16, wherein

the similarity degree between the current versions of the multiple Web pages is calculated by the following equation,

$\begin{array}{c}{\mathrm{CR}}_j=\sum _{}^{}\end{array}$

i = 1 W ⁢

⁢

[ cos ⁡

( V i , V j ) * ( 1 + ɛ

* W - R i SE W ) ]

where CR_j is a current similarity degree of a Web page j, cos is a cosine similarity, V_i is a vector of a current version of the Web page i, V_j is a vector of a current version of the Web page j, R_i^SE is the search engine ranking of the Web page i, ε

is the parameter, W is the total number of URLs as being an object to be re-ranked among the number of URLs of the search result obtained from the search engine.

18. The search system of claim 13, wherein

the super page producing part produces the super pages by referring to a Web archive part that memorizes multiple versions of Web pages that existed on the Internet in the past.

19. The search system of claim 13, wherein

the search result page is obtained by a searching process by a Web search engine.