Abstract :
Schema matching is a problem of finding correspondences, particularly equivalence relationships across schemas. The problem has a particular significance in integrating web repositories, as distributed databases over the web becomes increasingly popular. Most of the existing prototypes use schema level lexical information for schema matching. However, most of them perform rather poorly on real-world problems due to the abundance of abbreviations in real-world schemas. For example, none of the lexical matchers we tested would recommend a mapping of ´cnum´ to ´cid´, while ´customer number´ and ´customer ID´ are matching entities. In this work we propose a method for abbreviation expansion in schemas that facilitates lexical schema matching.
