Information recommendation device and information recommendation method

• US 8,108,376 B2
• Filed: 03/20/2009
• Issued: 01/31/2012
• Est. Priority Date: 03/28/2008
• Status: Active Grant

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation), Toshiba Digital Solutions Corporation (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Okamoto, Masayuki, Kikuchi, Masaaki
• Primary Examiner(s): Ali, Mohammad
• Assistant Examiner(s): Ruiz, Angelica
• Application Number:
  US12/407,827
• Publication Number:
  US 20090248678A1
• Time in Patent Office: 1,047 Days
• Field of Search: None
• US Class Current: 707/705
• CPC Class Codes: G06F 16/355 Class or cluster creation or modification

Abstract

A document set, and history documents including documents, etc., browsed by a user are input. The document set and the history documents are each analyzed to obtain characteristic vectors. A plurality of topic clusters and a plurality of sub-topic clusters are obtained by clustering the document set. A transition structure showing transitions of topics among the sub-topic clusters is generated, and a characteristic attribute is extracted from each topic cluster and each sub-topic cluster. An cluster-of-interest is extracted in comparison among characteristic vectors of the history documents and a characteristic vector of each document included in the document set, a sub-topic cluster having transition relations with the cluster-of-interest is obtained on the basis of a transition structure owned by the cluster-of-interest, and a document included in the sub-topic cluster is extracted as a recommended document to be presented together with the characteristic attribute.

Claims

1. An information recommendation device, comprising:

a document input unit which inputs a document set of which each document has date and time information within a specified time period;

a document analysis unit which obtains a plurality of characteristic vectors each including a plurality of keywords of vector elements by each keyword analyses of the document set or history documents including browsed documents or documents labeled by bookmark operations;

a clustering unit which obtains a plurality of topic clusters and a plurality of sub-topic clusters which are each composed of documents belonging to the same topic by clustering the document set;

a topic transition generation unit which generates a transition structure showing transitions of topics among the sub-topic clusters;

a characteristic attribute extraction unit which extracts a characteristic attribute of frequently included keyword from each topic cluster and each sub-topic cluster;

a cluster-of-interest extraction unit which extracts a cluster-of-interest equivalent to any one of the plurality of topic clusters or sub-topic clusters by similarity determination among the characteristic vectors of history documents and the characteristic vector of each document included in the document set;

a recommended document extraction unit which obtains a sub-topic cluster having transition relations with the cluster-of-interest on the basis of the transition structure owned by the cluster-of-interest, and extracts a document included in the sub-topic cluster as a recommended document; and

a recommended document presentation unit which presents the recommended document together with the characteristic attribute.

2. The device according to claim 1, further comprising:

a history input unit which inputs the history documents.

3. The device according to claim 1, further comprising:

a structuration determining unit which determines whether or not each of the plurality of topic clusters is configured to structuralize for further clustering each of the topic clusters into sub-topic clusters, and if it is configured to structuralize therefor, clusters the topic clusters, and controls the clustering unit so as to obtain the plurality of sub-topic clusters.

4. The device according to claim 1, further comprising:

a cluster structure storage unit which stores cluster structure information including the characteristic attribute and the transition structure.

5. The device according to claim 1, wherein the document set includes a document which has date and time information at the document itself and a document in which metadata accompanying the document has date and time information.

6. The device according to claim 1, wherein the clustering unit obtains an inner product value for a group of characteristic vectors corresponding to a set of arbitrary documents, and performs clustering on the basis of threshold determination of the inner product value.

7. The device according to claim 1, wherein the topic transition generation unit obtains similarity among documents included in the topic cluster and documents included in the sub-topic cluster, or relations among date and time information of documents included in the topic cluster and date and time information of documents included in the sub-topic cluster, and then, generates the transition structure.

8. The device according to claim 1, further comprising:

a preference structure storage unit which stores a preference structure based on the plurality of history documents, wherein

the clustering unit performs clustering for the plurality of history documents;

the preference structure storage unit stores the preference structure based on results of the clustering; and

the cluster-of-interest extraction unit extracts a cluster-of-interest in comparison between the preference structure and the cluster structure information.

9. The device according to claim 8, wherein the document analysis unit performs document analysis of each of the plurality of history documents, and weights preferences on the basis of the results of the document analyses.

10. An information recommendation method, comprising:

inputting a document set of which each document has date and time information within a specified time period;

obtaining a plurality of characteristic vectors which each include a plurality of keywords of vector elements by each keyword analyses of the document set or history documents including browsed documents or documents labeled by bookmark operations;

obtaining a plurality of topic clusters and a plurality of sub-topic clusters which are each composed of documents belonging to the same topic by clustering the document set;

generating a transition structure which shows transitions of topics among the sub-topic clusters;

extracting a characteristic attribute from each topic cluster and each sub-topic cluster;

extracting a cluster-of-interest equivalent to any one of the plurality of topic clusters or sub-topic clusters by similarity determination among characteristic vectors of the history documents and characteristic vector of each document included in the document set;

obtaining a sub-topic cluster which has transition relations with the cluster-of-interest on the basis of the transition structure owned by the cluster-of-interest, and extracting a document included in the sub-topic cluster as a recommended document; and

presenting the recommended document together with the characteristic attribute.

11. The method according to claim 10, further comprising:

inputting the history documents.

12. The device according to claim 10, further comprising:

determining whether or not each of the plurality of topic clusters is configured to structuralize for further clustering each of the topic clusters into sub-topic clusters, and if it is configured to structuralize therefor, clustering the topic clusters, and controlling the clustering so as to obtain the plurality of sub-topic clusters.

13. The method according to claim 10, further comprising:

storing cluster structure information which includes the characteristic attribute and the transition structure.

14. The method according to claim 10, wherein the document set includes a document which has date and time information within the document and a document in which metadata accompanying the document has date and time information.

15. The method according to claim 10, further comprising:

obtaining an inner product value for a group of characteristic vectors corresponding to a set of arbitrary documents, and clustering on the basis of threshold determination of the inner product value.

16. The method according to claim 10, further comprising:

obtaining similarity among documents included in the topic cluster and documents included in the sub-topic cluster, or relations among date and time information of documents included in the topic cluster and date and time information of documents included in the sub-topic cluster, and then, generating the transition structure.

17. The method according to claim 10, further comprising:

storing a preference structure based on the plurality of history documents;

clustering for the plurality of history documents;

storing the preference structure based on results of the clustering; and

extracting a cluster-of-interest in comparison between the preference structure and the cluster structure information.

18. The method according to claim 17, further comprising:

performing each document analysis of the plurality of history documents, and weighting preferences on the basis of the results of the document analyses.

19. A non-transitory computer readable medium including an information recommendation computer executable program, program, when executed by a computer, causes the computer to perform a method comprising:

the computer to input a document set of which each document has date and time information within a specified time period;

instructing the computer to obtain a plurality of characteristic vectors each including a plurality of keywords of vector elements by each keyword analyses of the document set or history documents including browsed documents or documents labeled by bookmark operations;

instructing the computer to obtain a plurality of topic clusters and a plurality of sub-topic clusters which are each composed of documents belonging to the same topic by clustering the document set;

instructing the computer to generate a transition structure showing transitions of topics among the sub-topic clusters;

instructing the computer to extract a characteristic attribute of frequently included keyword from each topic cluster and each sub-topic cluster;

instructing the computer to extract a cluster-of-interest equivalent to any one of the plurality of topic clusters or sub-topic clusters by similarity determination among the characteristic vectors of the history documents and the characteristic vector of each document included in the document set;

instructing the computer to obtain a sub-topic cluster having transition relations with the cluster-of-interest on the basis of the transition structure owned by the cluster-of-interest, and extract a document included in the sub-topic cluster as a recommended document; and

instructing the computer to present the recommended document together with the characteristic attribute.