Parsing Japanese Utterance with Self-repair

T. Harano, Y. Sagawa, and N. Sugie (Japan)


self-repair, Japanese utterance, parsing, morphological analysis


In spontaneous speech, speakers make many kinds of mistakes. Human listener can guess the intended meaning of the utterance even if it includes ill-formedness. However, the previous natural language processing systems including the spoken dialogue systems cannot cope with ill-formed utterances, and they have no flexibility like human. Conventional parsing systems also cannot cope with ill-formed sentences mostly. We aim to construct a parsing system that can cope with self-repairs, one of typical and frequent ill-formedness. In this paper, we propose the grammatical rule based on the analysis of characteristics of sentences including self-repairs, and evaluate its effectiveness.

Important Links:

Go Back