Background knowledge in ontology matching

Ontology matching is a critical operation in many well-known metadata intensive applications, such as data integration and peer-to-peer information sharing. Typically, heterogeneity in these applications is reduced in two steps: (i) matching ontologies to determine correspondences and (ii) executing correspondences according to an application needs (e.g., data translation). In this position statement paper we focus only on the first, i.e., matching step. In particular, we think of matching as an operation that takes two graph-like structures, such as lightweight ontologies [7], and produces a set of correspondences between the nodes of the graphs that correspond semantically to each other [8, 11]. Many diverse solutions of matching have been proposed so far, see [17, 19] for surveys. Also, recently this topic has been given a book account in [4]. It is worth noting that, on the one side, schema matching is usually performed with the help of techniques trying to guess the meaning encoded in the schemas. On the other side, ontology matching systems primarily try to exploit knowledge explicitly encoded in the ontologies. In real-world applications, various data and conceptual models usually have both well defined and obscure terms, and contexts in which they occur, therefore, solutions from both problems would be mutually beneficial [19]. Similar ideas of cross-fertilization among databases and artificial intelligence in the field of matching were also put forward in [16, 18]. Let us discuss one of the challenges in the matching area, which is the lack of background knowledge in matching tasks [9]. We believe that this challenge can be best tackled from the multi-disciplinary viewpoint, by building on top of the experiences in various communities, including databases, artificial intelligence and semantic web.