Share on Facebook Tweet on Twitter Share on LinkedIn Share by email
Multilingual Semantic Matching with OrdPath in Relational Systems

A Kumaran and Peter Carlin

Abstract

The volume of information in natural languages in electronic format is increasing exponentially. The demographics of users of information management systems are becoming increasingly multilingual. Together these trends create a requirement for information management systems to support processing of information in multiple natural languages seamlessly. Database systems, the backbones of information management, should support this requirement effectively and efficiently. Earlier research in this area had proposed multilingual operators for relational database systems, and discussed their implementation using existing database features.

In this paper, we specifically focus on the SemEQUAL operator, implementing a multilingual semantic matching predicate using WordNet. We explore the implementation of SemEQUAL using OrdPath, a positional representation for nodes of a hierarchy that is used successfully for supporting XML documents in relational systems. We propose the use of OrdPath to represent position within the Wordnet hierarchy, leveraging its ability to compute transitive closures efficiently. We show theoretically that an implementation using OrdPath will outperform those implementations proposed previously. Our initial experimental results confirm this analysis, and show that the OrdPath implementation performs significantly better. Further, since our technique is not specifically rooted to linguistic hierarchies, the same approach may benefit other applications that utilize alternative hierarchical ontologies.

Details

Publication typeInproceedings
Published inIEEE Data Engineering Bulletin: Multi-lingual Information Systems
PublisherIEEE
> Publications > Multilingual Semantic Matching with OrdPath in Relational Systems