A Robust Ontology-Based Method for Translating Natural Language Queries to Conceptual Graphs

A natural language interface is always desirable for a search system. While performance of machine translation for general texts with acceptable computational costs seems to reach a limit, narrowing down the domain to one of queries reduces the complexity and enables better translation correctness. This paper proposes a query translation method that is robust to ill-formed questions and exploits knowledge of an ontology for semantic search. It uses conceptual graphs as the target language for the translation. As a logical inter-lingua with smooth mapping to and from natural language, conceptual graphs simplify translation rules and can be easily converted to other formal query languages. Experiment results of the method on the TREC 2002 and TREC 2007 data sets are also presented and discussed.