Multi-schema matching based on clustering techniques

Abstract

Schema matching plays an important role in many database applications, such as ontology merging, data integration, data warehouse and dataspaces. The problem of schema matching is to find the semantic correspondence between attributes of schemas to be matched. In this paper, we propose multi-schema matching based on clustering techniques. Traditional matching techniques mainly address matching tasks between two attributes, namely pairwise-attribute correspondence. However, there exist lots of applications that require the semantic correspondence among multiple attributes. Thus, we will focus on matching multiple attributes, which is more difficult than pairwise-attribute correspondence. We employ the clustering techniques to solve the multi-schema matching problem. We use the well-known TFIDF weighting method to convert each attribute in schemas to a point in the vector space model. Then, these attributes can be partitioned into different clusters each of which has a specific semantics topic. Finally, the attributes partitioned into the same cluster are similar with higher confidence. We validate our approach with an experimental study, the results of which demonstrate that our approach is effective and has good performance.