Semi-automatic acquisition of domain-specific semantic structures

This paper describes a methodology for semi-automatic grammar induction from unannotated corpora belonging to a restricted domain. The grammar contains both semantic and syntactic structures, which are conducive towards language understanding. Our work aims to ameliorate the reliance of grammar development on expert handcrafting or the availability of annotated corpora. To strive for a reasonable model for real data, as well as portability across domain and languages, we adopt a statistical approach. Our approach is also amenable to the optional injection of prior knowledge to aid grammar induction, and subsequent hand editing for grammar refinement. This constitutes the semi-automatic nature of the approach. Experiments with the ATIS corpus showed positive results in semantic parsing, when compared to an entirely handcrafted grammar.