A FUZZY DOCUMENT-RETRIEVAL SYSTEM USING THE KEYWORD CONNECTION MATRIX AND A LEARNING-METHOD

We have already proposed a fuzzy document retrieval system using a keyword connection matrix to represent similarities between keywords. By using the keyword connection matrix, documents are graded according to their relevance to a user query. We have also proposed a learning method to modify the relationship values so as to reduce the difference between relationship values initially assigned using statistical information, and the user's evaluation. There were, however, two problems with our previous method: First, we restricted compound queries to logical OR; second, our method could not accept ambiguous user judgments in the learning process. We have extended our method to solve these two problems. In the new method, compound queries composed of keywords with AND, OR and/or NOT are processed, and the learning method has been modified to allow fuzzy judgements as well as compound queries. The new method has been implemented on a Unix workstation. Measurement of the recall and precision ratios has shown the effectiveness of our method. The averages for these ratios in several retrievals were 56% and 40% respectively, compared to 41% and 43% using the conventional crisp method. The effect of learning was also measured. The average of the recall and presision ratios were increased to 75% and 50% after 30 learning iterations.