Method and system for automatic objects classification

• US 8,275,765 B2
• Filed: 10/28/2009
• Issued: 09/25/2012
• Est. Priority Date: 10/30/2008
• Status: Active Grant

• Current Assignee: NEC Company Limited
• Original Assignee: NEC Company Limited
• Inventors: Li, Jianqiang, Meng, Xin, Zhao, Yu, Shi, Jingwei
• Primary Examiner(s): Pulliam, Christyann
• Assistant Examiner(s): Cao, Phuong Thao
• Application Number:
  US12/607,369
• Publication Number:
  US 20100114855A1
• Time in Patent Office: 1,063 Days
• Field of Search: 707/999.103, 707/722, 707/953, 707/E17.089, 707/737, 707/956
• US Class Current: 707/722
• CPC Class Codes: G06F 16/353 into predefined classes

Abstract

The present invention provides a method and system for automatic objects classification. The method comprises: acquiring a set of objects; classifying the objects based on query log to generate a first classification result; classifying the objects based on ontological information to generate a second classification result; and semantically fusing the first and second classification results to generate a final classification result. According to the present invention, compared with the prior arts, by semantically fusing the query log-based classification result and the ontology-based classification result, the accuracy and user-friendness of the object classification can be improved.

Claims

1. A method for objects classification, comprising:

acquiring, by at least one circuit, a set of objects;

classifying, by the at least one circuit, the objects based on a query log to generate a first classification result;

classifying, by the at least one circuit, the objects based on ontological information to generate a second classification result; and

semantically fusing, by the at least one circuit, the first and second classification results to generate a final classification result,

wherein the semantically fusing comprises adding a semantic element of one of the first classification result and the second classification result to the other of the first classification result and the second classification result, and

wherein the adding comprises;

tuning at least one of the first classification result according to the second classification result and the second classification result according to the first classification result, and

integrating, after the tuning, the first and second classification results to generate the final classification result.

2. The method according to claim 1, wherein the step of acquiring the set of objects comprises:

inputting a target query;

searching in an object base according to the target query; and

outputting search results as the set of objects to be classified.

3. The method according to claim 2, wherein said search in the object base employs a keyword-based information retrieval method.

4. The method according to claim 2, wherein the step of classifying the objects based on the query log comprises:

obtaining the query log;

extracting from the query log all the queries related to the target query;

clustering the extracted related queries and determining a cluster center for each cluster as an object category; and

classifying all the search results to the respective determined categories.

5. The method according to claim 2, wherein the step of classifying the objects based on ontological information comprises:

obtaining the ontological information;

annotating all the objects and the target query with reference to the ontological information;

generating object categories according to the semantic related concepts of the annotated target query in the ontological information; and

classifying all the search results to the respective generated categories.

6. The method according to claim 1, wherein:

the step of tuning comprises;

tuning the first classification result according to the second classification result, and

tuning the second classification result according to the first classification result; and

the step of integrating comprising integrating the tuned first and second classification results to generate the final classification result.

7. The method according to claim 6, wherein the step of tuning the first classification result comprises:

adding a semantic framework formed in the second classification result to the first classification result.

8. The method according to claim 6, wherein the step of tuning the second classification result comprises:

adding queries in the first classification result with high click frequency to the second classification result.

9. A system for objects classification, comprising:

an object acquiring unit comprising a circuitry for acquiring a set of objects;

a query log-based classification unit comprising a circuitry for classifying the objects based on a query log to generate a first classification result;

an ontological information-based classification unit comprising a circuitry for classifying the objects based on ontological information to generate a second classification result; and

a semantic fusing unit comprising a circuitry for semantically fusing the first and second classification results to generate a final classification result, wherein the semantically fusing comprises adding a semantic element of one of the first classification result and the second classification result to the other of the first classification result and the second classification result, and

wherein the semantic fusing unit tunes at least one of the first classification result according to the second classification result and the second classification result according to the first classification result, and integrates, after the tuning, the first and second classification results to generate the final classification result.

10. The system according to claim 9, further comprising:

an object base for storing the objects,

wherein the object acquiring unit comprises;

an input unit comprising a circuitry for inputting a target query; and

a search unit comprising a circuitry for searching in the object base according to the target query and outputting search results as the set of objects to be classified.

11. The system according to claim 10, wherein the query log-based classification unit comprises:

a query log obtaining unit comprising a circuitry for obtaining the query log;

a related query extraction unit comprising a circuitry for extracting from the query log all the queries related to the target query;

a cluster-based category learning unit comprising a circuitry for clustering the extracted related queries and determining a cluster center for each cluster as an object category; and

a classification unit comprising a circuitry for classifying all the search results to the respective determined categories.

12. The system according to claim 10, wherein the ontological information-based classification unit comprises:

an ontological information obtaining unit comprising a circuitry for obtaining the ontological information;

an object annotation unit comprising a circuitry for annotating all the objects that are search results with reference to the ontological information;

a query annotation unit comprising a circuitry for annotating the target query;

a category generation unit comprising a circuitry for generating object categories according to the semantic related concepts of the annotated target query in the ontological information; and

a classification unit comprising a circuitry for classifying all the search results to the respective generated categories.

13. The system according to claim 9, wherein the semantic fusing unit comprises:

a first tuning unit comprising a circuitry for tuning the first classification result according to the second classification result;

a second tuning unit comprising a circuitry for tuning the second classification result according to the first classification result; and

an integration unit comprising a circuitry for integrating the tuned first and second classification results to generate the final classification result.