Semantic interpretation of Dutch spoken dialogue (short paper)

Semantic interpretation involves the process of ‘translating’ natural language to a representation of its meaning. It could be understood as the task of mapping syntax to semantics, assuming that the syntactic relationships in an utterance correspond to functional relationships in the meaning representation. Relevant work in this area often uses techniques from machine translation and machine learning in the mapping from natural language to meaning-representation languages (e.g. [9, 7]). These approaches can be robust, and thus would be useful in dealing with large quantities of utterances, but require large amounts of annotated data. Since the syntax of natural language does not change much from domain to domain, an alternative way is to use the output of a wide-coverage syntactic parser as a basis for single, multiple, or even open-domain language processing. To obtain a sufficiently detailed semantic representation, the phrases in the parses should be linked with domain-specific knowledge concepts. In this paper, a syntactic parsing based system for the semantic interpretation of Dutch spoken language is presented and evaluated.