Data retrieval method, system and program product

• US 7,219,089 B2
• Filed: 09/24/2003
• Issued: 05/15/2007
• Est. Priority Date: 10/04/2002
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Kobayashi, Mei, Aono, Masaki
• Primary Examiner(s): Kindred; Alford
• Assistant Examiner(s): NGUYEN, MERILYN P
• Application Number:
  US10/669,799
• Publication Number:
  US 20040068493A1
• Time in Patent Office: 1,329 Days
• Field of Search: 707/3, 707/4, 707/101, 707/1, 707/5, 704/4, 704/8, 704/9, 715/501.5
• US Class Current: 1/1
• CPC Class Codes: G06F 16/358 Browsing; Visualisation therefor
• CPC Class Codes: Y10S 707/99931 Database or file accessing
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99934 Query formulation, input preparation, or translation
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

A data retrieval system having a database storing data as a vector that has been digitized based on a keyword. The system has capability of generating a residual vector from the data to compute a covariance matrix and an eigenvector of the covariance matrix. The system further has the capability for generating and storing a set of basic vectors from the set of the computed eigenvectors, and for reading out the data and at least one of the eigenvectors from a memory. This is utilized to compute a contribution of the eigenvector to the data, and for contracting or enlarging a residual vector to store. The system selects a keyword to be used for labeling according to a similarity between the stored basic vector and the data, and a weight on the similarity so as to store the keyword in a memory.

Claims

1. A data retrieval system for causing a computer to retrieve data being stored in a database, said retrieval system comprising:

a database storing data as a vector digitized based on a keyword;

a means for generating a residual vector from said data, said residual vector corresponding to a vector in which an element corresponding to a contribution component in a direction of a basic vector calculated is subtracted from a previously obtained residual vector, the basic vector and the residual vector a newly generated residual vector lie in an orthogonal relationship, to compute and store a covariance matrix and an eigenvector of said covariance matrix, and for generating and storing a set of basic vectors from a set of computed eigenvectors;

a means for reading out said data and at least one of said eigenvectors from the database, and for computing and storing a contribution vector of said eigenvector to said data, and for contracting or enlarging a residual vector by reading out said contribution vector to compute and store a new eigenvector; and

a means for selecting a keyword to be used for labeling clusters according to a similarity between said stored basic vector and said data, and a weight on said similarity so as to store the keyword in the database;

means for classifying data into clusters of documents having the same or similar keywords and depending on a similarity between the stored basic vector and the data; and

a means for outputing cluster data of a cluster to a graphical user interface system for displaying the cluster data.

2. The data retrieval system according to claim 1, comprising a means for making said basic vectors orthogonal.

3. The data retrieval system according to claim 1, wherein said means for selecting the keyword to be used for labeling clusters to store the keyword in the database further comprises a means for determining the weight on said similarity to said keyword and a means for storing a certain number of keywords in a descending order in the database in connection with said weight.

4. A data retrieval method for causing a computer to retrieve data stored in a database, said data retrieval method comprising the steps of:

reading out data from a database storing data as a vector digitized based on a keyword;

computing and storing a covariance matrix and an eigenvector of said covariance matrix, using said data;

generating and storing a set of basic vectors from a set of said computed eigenvectors;

reading out said data and at least one eigenvector from the database, and computing and storing a contribution vector of said eigenvector to said data; and

computing a residual vector from said data and said eigenvector, and contracting or enlarging a residual vector by reading out said contribution vector to compute and store a new eigenvector; and

classifying data into clusters of documents having the same or similar keywords and depending on a similarity between the stored basic vector and the data; and

outputting cluster data of a cluster to a graphical user interface system for displaying the cluster data.

5. The data retrieval method according to claim 4, further comprising a step of selecting a keyword to be used for labeling clusters from a similarity between said stored basic vector and said data and a weight on said similarity to store the keyword in the database.

6. The data retrieval method according to claim 5, wherein said step of selecting the keyword to be used for labeling clusters further comprises a step of determining the weight on said similarity to said keyword and storing a certain number of keywords in a descending order in the database in connection with said weight.

7. A computer executable program product comprising a computer readable media having computer executable program thereon for implementing a data retrieval method for causing a computer to retrieve data stored in a database, said program comprising the steps of:

reading out data from a database storing data as a vector digitized based on a keyword;

computing and storing a covariance matrix and an eigenvector of said covariance matrix, using said data;

generating and storing a set of basic vectors from a set of said computed eigenvectors;

reading out said data and at least one eigenvector from the database, and computing and storing a contribution vector of said eigenvector to said data, and computing a residual vector from said data and said eigenvector, and contracting or enlarging a residual vector by reading out said contribution vector to compute and store a new eigenvector in the database; and

classifying data into clusters of documents having the same or similar keywords and depending on a similarity between the stored basic vector and the data; and

outputting cluster data of a cluster to a graphical user interface system for displaying the cluster data.

8. The program product according to claim 7, said program further comprising a step of selecting a keyword to be used for labeling clusters from a similarity between said stored basic vector and said data and a weight on said similarity to store the keyword in the database.

9. The program product according to claim 7, said program further comprising a step of making said basic vectors orthogonal before computing said residual vector.