System and method for disambiguating entities in a web page search

• US 7,386,545 B2
• Filed: 03/31/2005
• Issued: 06/10/2008
• Est. Priority Date: 03/31/2005
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: International Business Machines Corporation
• Inventors: Ashutosh, Garg, Guha, Ramanathan
• Primary Examiner(s): MIZRAHI, DIANE D
• Application Number:
  US11/095,273
• Publication Number:
  US 20060224565A1
• Time in Patent Office: 1,167 Days
• Field of Search: 707 3- 6
• US Class Current: 1/1
• CPC Class Codes: G06F 16/35 Clustering; Classification
• 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 system and method of disambiguating entities in a computerized web search includes identifying a set of potential meanings for an entity; retrieving at least one web page having descriptions referencing the entity; establishing a base web page having a selected context for the entity; attributing dimensions of a vector space attributed to domains in the retrieved web page; and computing a probability of similarity between the referenced entity in the retrieved web page and the entity in the base web page. The method includes corresponding a similarity measure between the dimensions of the vector space attributed to domains in the retrieved web page and a likelihood of the retrieved web page referring to the entity in the base web page. The method further includes ranking web pages based on the computed probability of similarity.

Claims

1. A method of disambiguating entities in a computerized web search, said method comprising:

identifying a set of potential meanings for an entity, wherein said entity comprises any of a word and phrase;

retrieving at least one retrieved web page comprising descriptions referencing said entity;

establishing a base web page comprising a selected meaning of said potential meanings for said entity;

attributing dimensions of a vector space attributed to domains in said retrieved web page;

computing a probability of similarity between said descriptions in said retrieved web page and said entity in said base web page,

wherein said computing of said probability of similarity comprises corresponding a similarity measure between said dimensions of said vector space attributed to domains in said retrieved web page and a likelihood of said retrieved web page referring to said entity in said base web page; and

reporting said probability of said similarity to a user.

2. The method of claim 1, all the limitations of which are incorporated herein by reference, wherein said dimensions comprise any of an object and a word.

3. The method of claim 1, all the limitations of which are incorporated herein by reference, wherein said at least one web page comprises a plurality of web pages, wherein each web page comprises a description of said entity that matches a search entity chosen for a search query.

4. The method of claim 3, all the limitations of which are incorporated herein by reference, wherein said base web page is selected from one of said web pages.

5. The method of claim 1, all the limitations of which are incorporated herein by reference, further comprising ranking web pages based on the computed probability of similarity.

6. A program storage device readable by computer, tangibly embodying a program of instructions executable by said computer to perform a method of disambiguating entities in a computerized web search, said method comprising:

identifying a set of potential meanings for an entity, wherein said entity comprises any of a word and phrase;

retrieving at least one retrieved web page comprising descriptions referencing said entity;

establishing a base web page comprising a selected meaning of said potential meanings for said entity;

attributing dimensions of a vector space attributed to domains in said retrieved web page;

computing a probability of similarity between said descriptions in said retrieved web page and said entity in said base web page,

wherein said computing of said probability of similarity comprises corresponding a similarity measure between said dimensions of said vector space attributed to domains in said retrieved web page and a likelihood of said retrieved web page referring to said entity in said base web page; and

reporting said probability of said similarity to a user.

7. The program storage device of claim 6, all the limitations of which are incorporated herein by reference, wherein in said method, said dimensions comprise any of an object and a word.

8. The program storage device of claim 6, all the limitations of which are incorporated herein by reference, wherein in said method, said at least one web page comprises a plurality of web pages, wherein each web page comprises a description of said entity that matches a search entity chosen for a search query, and wherein said base web page is selected from one of said web pages.

9. The program storage device of claim 6, all the limitations of which are incorporated herein by reference, wherein said method further comprises ranking web pages based on the computed probability of similarity.

10. A system for disambiguating entities in a computerized web search, said system comprising:

a user interface adapted to identify a set of potential meanings for an entity, wherein said entity comprises any of a word and phrase;

a search engine connected to said user interface and adapted to retrieve at least one retrieved web page comprising descriptions referencing said entity;

a processor adapted to;

establish a base web page comprising a selected meaning of said potential meanings for said entity;

attribute dimensions of a vector space attributed to domains in said retrieved web page; and

compute a probability of similarity between said descriptions in said retrieved web page and said entity in said base web page,

wherein said computing of said probability of similarity comprises corresponding a similarity measure between said dimensions of said vector space attributed to domains in said retrieved web page and a likelihood of said retrieved web page referring to said entity in said base web page; and

a display adapted to report said probability of said similarity to a user.

11. The system of claim 10, all the limitations of which are incorporated herein by reference, wherein said dimensions comprise any of an object and a word.

12. The system of claim 10, all the limitations of which are incorporated herein by reference, wherein said at least one web page comprises a plurality of web pages, wherein each web page comprises a description of said entity that matches a search entity chosen for a search query.

13. The system of claim 12, all the limitations of which are incorporated herein by reference, wherein said base web page is selected from one of said web pages.

14. The system of claim 10, all the limitations of which are incorporated herein by reference, wherein said search engine is adapted to rank web pages based on the computed probability of similarity.