An Ontology-based Semantic Clustering Algorithm for Accounting Text

Abstract

The feature selection and semantic similarity computing between texts are essential components of accounting text clustering. In the past, several approaches for generic text feature selection and similarity computing by exploiting different measures (vector space model, words frequency, thesauri, domain corpora, etc.) have been proposed. However, accounting field is different from general field. Accounting has its own concepts and rules. These generic methods are not so suitable for accounting text clustering. In this paper, a novel accounting ontology-based feature selection and similarity computing algorithm for accounting text is proposed. Firstly, characterizing the accounting texts, we get a terms vector. Secondly, terms vector is mapped into concept of accounting ontology and converted into concept vector. Based on the structure of concept, the semantic similarity between texts is computed. Then, trough an improved clustering method, accounting texts are clustered effectively. The experiments results imply that our proposal outperforms most of the previous measures as well as eliminates some of their limitations.