To improve the recognition accuracy for spontaneous conversational speech, we collected a corpus to study how spontaneous conversational speech differs from read style speech. The corpus consists of two parts: 1) spontaneous conversational speech and 2) read speech with the same word transcriptions as the conversational speech. In word and phone recognition experiments, it was confirmed that, for the Japanese language, the recognition of spontaneous speech is harder than that of read speech. By comparing of recognition results, we found that, both in the occurrence of errors appearing with speaking style changes, and in the types of pronunciation variants, there are differences that depend on the linguistic categories that misrecognized words belong to. We confirmed that linguistic categories also affect pronunciation variants that deteriorate the recognition accuracy.
Cite as: Nakajima, H., Hirano, I., Sagisaka, Y., Shirai, K. (2001) Pronunciation variant analysis using speaking style parallel corpus. Proc. 7th European Conference on Speech Communication and Technology (Eurospeech 2001), 65-68, doi: 10.21437/Eurospeech.2001-15
@inproceedings{nakajima01_eurospeech,
author={Hideharu Nakajima and Izumi Hirano and Yoshinori Sagisaka and Katsuhiko Shirai},
title={{Pronunciation variant analysis using speaking style parallel corpus}},
year=2001,
booktitle={Proc. 7th European Conference on Speech Communication and Technology (Eurospeech 2001)},
pages={65--68},
doi={10.21437/Eurospeech.2001-15}
}