The CPK NLP suite for spoken language understanding

This paper describes a number of freely available tools for implementing and running spoken language understanding systems. Unlike other free tools (e.g. the CSLU toolkit), the main emphasis is on spoken language understanding (syntactic/semantic parsing, generation of language models for recognition etc.). The suite supports (reads and/or writes) a number of grammar formats defined for speech recognisers like Entropic’s GrapHvite (HTK standard lattice) as well as common unification grammar formats used in NLP: The American PATRII and the European EUROTRA (Augmented Phrase Structure Grammar) format. The open architecture is completed by a general API allowing simple interfacing to dialogue management and speech recognition. The suite, implemented in C++ (partly C, Lex, Yacc), can compile and run on any machine under any OS having a 32 bit C++ compiler, Flex/Bison or Lex/Yacc. The suite can be downloaded at http://www.cpk.auc.dk/~tb/nlpsuite.