Interview-style-based Method of Collecting Spontaneous Speech Corpus for Car Navigation Systems

Kazuaki SHIMA  Takeshi HOMMA  Rintaro IKESHITA  Hiroaki KOKUBO  Yasunari OBUCHI  Jinhua SHE  

D - Abstracts of IEICE TRANSACTIONS on Information and Systems (Japanese Edition)   Vol.J101-D   No.2   pp.446-455
Publication Date: 2018/02/01
Online ISSN: 1881-0225
Type of Manuscript: PAPER
interview,  language understanding,  corpus,  morpheme,  navigation system,  spontaneous speech,  

Full Text(in Japanese): PDF(1010.7KB)
>>Buy this Article

This paper presents a new interview-based method of collecting utterance corpus for the development of modules of spoken language understanding (SLU). The method collects two types of utterances: (1) simple utterances collected from the use of current voice-enabled products, and (2) natural utterances collected from improved voice-enabled products, which are able to have a much more natural conversation with a human being than current products do. The morphological analysis on collected corpus show that the corpus covered a wider variety of expressions than sentences of log data acquired by current products do, and the corpus improved SLU accuracy of current products by utilizing it as training data of an SLU module.