Clustering fuzzy expected value system

• US 5,414,797 A
• Filed: 07/28/1992
• Issued: 05/09/1995
• Est. Priority Date: 05/16/1991
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Vassiliadis, Stamatis, Kobrosly, Walid, Triantafyllos, George
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US07/920,723
• Time in Patent Office: 1,015 Days
• Field of Search: 395/600, 395/934, 395/900, 395/51, 395/61
• US Class Current: 706/46
• CPC Class Codes: G06F 16/2468 Fuzzy queries
• CPC Class Codes: G06F 16/3334 Selection or weighting of terms from queries, including natural language queries
• CPC Class Codes: G06F 16/334 Query execution G06F16/335 takes precedence
• CPC Class Codes: G06F 18/232 Non-hierarchical techniques
• CPC Class Codes: G06F 8/20 Software design
• CPC Class Codes: G06N 5/048 Fuzzy inferencing
• CPC Class Codes: G06N 7/02 using fuzzy logic computing arrangements based on biological models G06N3/00; computing arrangements using knowledge-based models G06N5/00
• CPC Class Codes: Y10S 706/90 Fuzzy logic
• CPC Class Codes: Y10S 706/934 Information retrieval or Information management
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

A system provides a tool for computing the most typical fuzzy expected value of a membership function in a fuzzy set. The clustering fuzzy expected value system is used in a question answering system. CFEV is computed by the tool is based on grouping of individual responses, that meet certain criteria, to clusters. Each cluster is considered a "super response" and contributes to the result proportionally to its relative size and the difference in opinion from the mean of the entire sample. In so doing, CFEV represents the opinion of the majority of the population, but it also respects the opinion of the minority. A comparison is made with existed tools such as the FEV and the WFEV and the advantages of CFEV are demonstrated by examples for cases where other methods fail to perform.

Claims

1. A computer system for processing a database having language text records to evaluate their relevance to a predetermined subject of interest, the system comprising:

processors for processing instructions and the text records;

database storage means for storing the text records in the database;

a unique word generator for generating and storing a list of unique words contained in the database;

a relevant word generator for generating and storing a list of relevant words including a system user'"'"'s selections from the stored list of unique words;

a modified database generator for generating and storing a modified database, the modified database including the list of relevant words and synonyms, if any, associated with each word in the list of relevant words, whereinafter the unique word generator generates a list of unique words of the modified database;

a relevant word table including the list of unique words of the modified database; and

means for calculating and storing confidence values associated with each word in the relevant word table, each of the confidence values based on input values from a plurality of personnel other than the system user that reflects their perceptions of a relevance of said each word in the relevant word table to the subject of interest;

said means for calculating and storing confidence values including a clustering fuzzy expected value system for determining a membership grade for said each word in the relevant word table;

said clustering fuzzy expected value system determining said membership grade for a word in the relevant word table including grouping the confidence values for said word in the relevant word table into a plurality of clusters according to a predetermined formula, determining a mean of all the confidence values for said word in the relevant word table, and determining a plurality of mean confidence values each associated with one of said plurality of clusters of said confidence values.

2. The system according to claim 1, wherein the clustering fuzzy expected value system determines the mean of all the confidence values for said word in the relevant word table according to ##EQU19## determines the plurality of mean confidence values according to ##EQU20## and determines the membership grade according to ##EQU21## where i designates a cluster, a_ij is the number of confidence values in the cluster having value x_ij, m is the number of clusters, x is a parameter based on the field of use of the CFEV, N_i is the number of confidence values in cluster i, and N is the number of all the confidence value.

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

a clustering fuzzy evaluator for calculating and storing a clustered fuzzy expected value for each record in the database, wherein the membership grades are grouped into clusters according to the predetermined formula, and wherein the clustered fuzzy expected value for each record in the database is calculated based on a mean of the membership grades for all words in the list of relevant words and on each mean of the membership grades in each of said clusters.

4. The system according to claim 3, wherein the clustering fuzzy evaluator calculates the clustered fuzzy expected value for a record containing zero words to be equal to zero.

5. The system according to claim 3, wherein the clustering fuzzy evaluator calculates the clustered fuzzy expected value for a record containing one word to be equal to the membership grade for the one word.

6. The system according to claim 3, wherein the clustering fuzzy evaluator calculates the clustered fuzzy expected value for a record containing two words according to the formula ##EQU22## where w_i and w_j are membership grades for the two words;

k is a constant greater than zero;

l is a value greater than zero and less than one; and

c is equal to (l-m)/m where m is a mean of w_i and w_j.

7. The system according to claim 3, wherein the clustering fuzzy evaluator calculates the clustered fuzzy expected value for a record containing three or more words according to ##EQU23## where W_A is the mean of membership grades for all words in the record containing three or more words and is calculated according to ##EQU24## where W_Ai is the mean membership grade associated with one of the plurality of clusters of said membership grades and is calculated according to ##EQU25## and where N equals the number of words in the record containing three or more words, i designates a cluster, a_ij is the number of words in the cluster having membership grade x_ij, m is the number of clusters, n is the number of words in a cluster, x is a parameter based on the field of use of the CFEV, N_i is the number of words in cluster i.

8. The system according to claim 3, wherein the system further comprises means for retrieving a record having a fuzzy expected value greater than a predetermined value.

9. A method for evaluating a relevance of a database having text records to a predetermined subject of interest, the method comprising computer executed steps of:

compiling and storing a list of unique words contained in the database having text records;

extracting and storing from the list of unique words a list of relevant words selected by a user of the method;

storing confidence values furnished by personnel other than the user who are familiar with the subject of interest that reflect their perceptions of a relevance of each word in the list of relevant words to the subject of interest; and

calculating and storing a membership grade for each word in the list of relevant words, including grouping the confidence values for one word of the list of relevant words into a plurality of clusters according to a predetermined formula, determining a mean of the confidence values for said one word, and determining for each of said clusters a mean of the confidence values in each of said clusters.

10. The method according to claim 9, wherein the step of determining a mean of the confidence values for said one word is according to ##EQU26## the step of determining a mean of the confidence values in each of said clusters is according to ##EQU27## and the step of calculating a membership grade is according to ##EQU28## where i designates a cluster, a_ij is the number of confidence values in the cluster having value x_ij, m is the number of clusters, x is a parameter based on a field of use of the method, N_i is the number of confidence values in cluster i, and N is the number of all the confidence values for said one word.

11. The method according to claim 9, further comprising a computer executed step of calculating a clustered fuzzy expected joined confidence]value of each text record of the database based on fuzzy logic that incorporates the membership grade of each word in the text record.

12. The method according to claim 11, further comprising a computer executed step of retrieving a text record of the database having a clustered fuzzy expected value within a predetermined range.

13. The method according to claim 11, wherein the step of calculating a clustered fuzzy expected value of each text record is based on the formula ##EQU29## for records containing two words, where w_i and w_j are membership grades of the two words;

k is a constant greater than zero;

l is a value greater than zero and less than one; and

c is equal to (l-m)/m where m is a mean of w_i and w_j.