Soundness of Schema Matching Methods

Abstract

One of the key challenges in the development of open semantic-based systems is enabling the exchange of meaningful information across applications which may use autonomously developed schemata. One of the typical solutions for that problem is the definition of a mapping between pairs of schemas, namely a set of point-to-point relations between the elements of different schemas. A lot of (semi-)automatic methods for generating such mappings have been proposed. In this paper we provide a preliminary investigation on the notion of correctness for schema matching methods. In particular we define different notions of soundness, strictly depending on what dimension (syntactic, semantic, pragmatic) of the language the mappings are defined on. Finally, we discuss some preliminary conditions under which a two different notions of soundness (semantic and pragmatic) can be related.

OntoMatch: A monotonically improving schema matching system for autonomous data integration

“…For these two terms, editDistance(Leave,Departing) ≤ synonym(Leave,Departing) will always hold. But systems such as [20,2,3,9] apply matching algorithms without any implicit order, and do not take the position that application of edit distance matching is not warranted at all when a synonym matcher succeeds. Not only that it is inappropriate, we need not spend the time to blindly compute edit distance only to ignore it at the end by adjusting the weight to zero and thereby incur wasteful computational cost without necessarily leading to any improvement in the matching quality [10].…”

Advances in Ontology Matching

“…Alignments were evaluated in [6] using semantic soundness and completeness. They start from some representation language L (e.g., a Description Logic language [4]).…”

“…Imperfection in ontology matching has been discussed both in [54] and in [6]. The former argues for the need "to incorporate inaccurate mappings [correspondences] and handle uncertainty about mappings.…”

“…mappings may be inaccurate [since] the mapping language is too restricted to express more accurate mappings." [6] went even further, arguing philosophically that even if two ontologies fully agree on the semantics and the language is rich enough, ontologies may still not convey the same meaning, due to some hidden semantics, beyond the scope of the ontologies. A similar argument was provided in [59] in the context of relational databases: "the syntactic representation of schemas and data do not completely convey the semantics of different databases."…”

A Framework to Support Interoperability among Semantic Resources

“…There are many different possible ways to compute such similarity with various methods designed in the context of data analysis, machine learning, language engineering, statistics, or knowledge representation. FUNSIEC relies on semantic methods [3] to deal with the problem. The basic assumption behind semantic methods is that they aim at discovering relations between (pairs of) entities belonging to different schemata based on the meaning of the two entities (their semantics!).…”