Ontology-Based Approach for Automated Issue Classification in an Issue Tracking System

Abstract

A new ontology-based approach is presented for automated issue classification in an Issue Tracking System. An important and novel aspect of this approach is that the proposed fuzzy classification method does not require a training set, which is in contrast to the traditional statistical and probabilistic methods that require a set of pre-classified documents in order to train the classifier. Ontology is used as semantic knowledge representation for concept mapping and synonym extraction. The proposed approach is divided into three phases; Intelligent Pre-processing, Fuzzy Membership Calculation for the overall issue and finally Issue classification. Intelligent pre-processing is a new technique that does not blindly remove all stop words like traditional techniques but uses a concept tokenization rather than keyword tokenization. A set of ontologies were constructed for each domain of discourse for the Issue Tracking System (ITS) database. These ontologies are used to identify concepts and their relations for each domain that is appeared in the issue being processed. A Semantic Issue Tracking System - SIST - has been developed for applying this approach. The issue may be classified to more than one team with a degree of relevance depending on its semantic. Experiments show significant enhancement in issue classification over traditional Issue Tracking Systems.