SemEQUAL: Multilingual Semantic Matching in Relational Systems

In an increasingly multilingual world, it is critical that information management tools organically support the simultaneous use of multiple natural languages. A pre-requisite for efficiently achieving this goal is that the underlying database engines must provide seamless matching of text data across languages. We propose here SemEQUAL, a new SQL functionality for semantic matching of multilingual attribute data. Our current implementation defines matches based on the standard WordNet linguistic ontologies. A performance evaluation of SemEQUAL, implemented using standard SQL:1999 features on a suite of commercial database systems indicates unacceptably slow response times. However, by tuning the schema and index choices to match typical linguistic features, we show that the performance can be improved to a level commensurate with online user interaction.