Fuzzy Ontology Based Document Feature Vector Modification Using Fuzzy Tree Transducer

Abstract

Recently, an emphasis has been placed on the content based Information Retrieval Systems (IRS). Finding documents based on content similarity using background knowledge is becoming an increasingly important task. This paper aims for two main tasks in order to high quality document retrieval; first, we present our formulation of fuzzy ontology and then, by this formulation, propose a method which uses two functions for manipulating document feature vector. We describe encoding a fuzzy ontology into a Fuzzy Tree Transducer (FIT) and then, define two simple functions for applying attained FIT on document feature vector. By using the first function, elements of document feature vector are modified to reduce distance between current document and relevant documents in the vector space. This reduction is so important for categorization of documents in index repository of IRSs. The second function uses the injection of relevant context into query term. The injected context causes relocation of query term in vector space, and reduces its distance from some semantically related documents.