Document and information retrieval method and apparatus

• US 20040078363A1
• Filed: 09/02/2003
• Published: 04/22/2004
• Est. Priority Date: 03/02/2001
• Status: Active Grant

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Kawatani, Takahiko
• Granted Patent: US 7,194,461 B2
• Time in Patent Office: Days
• US Class Current: 707/3
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: Y10S 707/99934 Query formulation, input preparation, or translation
• CPC Class Codes: Y10S 707/99945 Object-oriented database structure processing

Abstract

A data procassing unit in programmed to retrieve a document item and/or an information item from a plurality of document items and/or information items. Each of the items is identified by an index vector. The retrieval (15) is in response to a query (11) including plural query terms related to each other by Boolean logic. The program causes the data processing unit to transform the query into vector form through matrix calculations (12) and to measure the similarities of the item index vectors and the vector form of the query to determine which of the items correspond with the query.

Claims

1. A method of retrieving a desired document item or information item from a plurality of document items and/or information items, items being identified by item index vectors, the method being performed in response to a query including plural query terms, the query terms being related to each other in Boolean logic form, the method comprising transforming the query terms in Boolean logic form into a transformed vector form, and retrieving the desired item information in response to similarity measurements of the transformed vector form of the query terms and the item index vectors.

2. The method of claim 1 wherein the transforming step includes:

(a) calculating a square sum matrix by the using a plurality of first index vectors, the first index vectors having components indicting the presence or absence of each of the plural query terms, the square matrix being calculated on the basis of all the first index vectors which are generated from said query; and

(b) calculating eigenvectors and eigenvalues of the square sum matrix.

3. The method of claim 2 wherein the transforming step further includes selecting basis vectors from the eigenvectors, the basis vectors constituting a subspace.

4. The method of claim 3 wherein the similarity measurements include calculating inner products between the first index vectors and the basis vectors, and weighted coefficients employing the eigenvalues.

5. The method of claim 4 wherein the retrieving step includes comparing, the similarity measurement with a predetermined threshold to determine whether or not each of said items relevant to said query.

6. The method of claim 5 further including varying the weighted coefficients and the threshold to maximize a predetermined evaluation measure.

7. The method of claim 2, further comprising the step of modifying said square sum matrix in response to at least one of (a) the first index vector being judged as being irrelevant in spite of being relevant to said query and (b) the first index vector being judged to be relevant in spite of being irrelevant to said query.

8. The method of claim 2 or 7, further comprising modifying said square sum matrix in response to at least one of (a) a user deciding that a retrieval item agrees with the query and (b) deciding that a retrieval item does not agree with the query.

9. The method of claim 2 or 7, further comprising modifying said square sum matrix on the basis of another item index vector having a component that indicates the presence or absence of each index term included in each of said items, and which corresponds to (a) the document judged by a user to agree with the user'"'"'s request, or (b) the document judged by the user to disagree with the user'"'"'s request.

10. The method of claim 9, further comprising:

(a) calculating a feedback vector by the using an average vector of the another index vectors which are included in said item judged by said user to agree with the user'"'"'s request, or an average vector of the another index vectors which are included in said item judged by said user to disagree;

(b) measuring a second similarity between each of the item index vectors and the feedback vector;

(c) calculating a third similarity by using said similarity measurement of claim 4 and the second similarity; and

(d) judging whether or not each item is relevant to said query by comparing the third similarity with a predetermined threshold.

11. The method of claim 2, wherein said square sum matrix is calculated in accordance with:

12. Apparatus for retrieving a document item and/or information item from a plurality of document items and/or information items, each of the items being identified by an item index vector, the apparatus, comprising:

(a) an input unit for accepting a query including plural query terms related to each other by Boolean logic;

(b) a data processing unit connected to be responsive to the input unit, the data processing unit being programmed to be responsive to the query for transforming the query into vector form, and for measuring the similarities of the item index vectors and the vector form of the query to determine which of the times correspond with the query, and an output device connected to be responsive to the data processing unit for providing an indication of the determination of which of the items correspond with the query.

13. The apparatus of claim 12 wherein the data processing unit is programmed to:

(a) calculate a square sum matrix by using (i) a first index vector having a component indicating the presence or absence of each query term included in said query, and (ii) all of the first index-vectors which documents relevant to said query can take;

(b) Galculate eigenvectors and eigenvalues of the square sum matrix;

(c) select basis vectors which constitute a subspace, from the eigenvectors;

(d) measure a first similarity between each of the items and said query on the basis of inner products between the item index vectors and the basis vectors, and weighted coefficients employing the eigenvalues;

(e) compare the first similarity with a predetermined threshold so as to determine whether or not each of said items is relevant to said query; and

(f) output a result of the judgement to the output unit.

14. The apparatus of claim 13, wherein the data processing unit is programmed for modifying said square sum matrix in response to at least one of (a) the first index vector being judged irrelevant in spite of being relevant to said query and (b) the first index vector being judged relevant in spite of being irrelevant to said query.

15. A program for controlling a data processing unit used to assist in retrieving a document item and/or an information item from a plurality of document items and/or information items, each of the items being identified by an item index vector, the retrieval being in response to a query including plural query terms related to each other by Boolean logic, the program being arranged for causing the data processing unit to transform the query into vector form and to measure the similarities of the item index vectors and the vector form of the query to determine which of the items correspond with the query.

16. The program of claim 15 wherein the program is arranged for causing the data processing unit to modify said square sum in response to at least one of (a) the first index vector being judged irrelevant in spite of being relevant to said query and (2) the first index vector being judged relevant in spite of being irrelevant to said query.

17. The program of claim 16 wherein the program is arranged for causing the data processing unit to modify said square sum matrix in response to at least one of (a) the first index vector being judged irrelevant in spite of being relevant to said query, and (b) the first index vector being judged relevant in spite of being irrelevant to said query.