Lexical Knowledge Extraction: an Effective Approach to Schema and Ontology Matching

This paper’s aim is to examine what role Lexical Knowledge Extraction plays in data integration as well as ontology engineering.Data integration is the problem of combining data residing at distributed heterogeneous sources, and providing the user with a unified view of these data; a common and important scenario in data integration are structured or semi-structure data sources described by a schema.Ontology engineering is a subfield of knowledge engineering that studies the methodologies for building and maintaining ontologies. Ontology engineering offers a direction towards solving the interoperability problems brought about by semantic obstacles, such as the obstacles related to the definitions of business terms and software classes. In these contexts where users are confronted with heterogeneous information it is crucial the support of matching techniques. Matching techniques aim at finding correspondences between semantically related entities of different schemata/ontologies.Several matching techniques have been proposed in the literature based on different approaches, often derived from other fields, such as text similarity, graph comparison and machine learning.This paper proposes a matching technique based on Lexical Knowledge Extraction: first, an Automatic Lexical Annotation of schemata/ontologies is performed, then lexical relationships are extracted based on such annotations.Lexical Annotation is a piece of information added in a document (book, online record, video, or other data), that refers to a semantic resource such as WordNet. Each annotation has the property to own one or more lexical descriptions. Lexical annotation is performed by the Probabilistic Word Sense Disambiguation (PWSD) method that combines several disambiguation algorithms.Our hypothesis is that performing lexical annotation of elements (e.g. classes and properties/attributes) of schemata/ontologies makes the system able to automatically extract the lexical knowledge that is implicit in a schema/ontology and then to derive lexical relationships between the elements of a schema/ontology or among elements of different schemata/ontologies.The effectiveness of the method presented in this paper has been proven within the data integration system MOMIS.