Semantic Decomposition of Sentences in the System Supporting Flight Services

Abstract

The paper describes the semantic part of the linguistic analysis of a recognized word hypothesis lattice, as implemented in an information dialogue system, capable of recognizing and understanding Slovenian speech. It describes the new Slovenian speech database from a semantic point of view. Semantic analysis is performed in three different steps: first, parsing of complex temporal expressions; second, parsing of simple noun phrases: this method is used for the departure/arrival location determination and in the third step we are looking for keywords defining the intended question . We propose some templates dealing with timetables, departure and arrival times, airline companies, etc. The different templates compete on each utterance. Finally, the template with the best score generates a database query. Some experimental results of the semantic decomposition are presented.