The branching factor and the perplexity have been used to measure the complexity of speech recognition task. In this paper, we state their disadvantages and propose that we should use the "language (task) entropy". We found the relationship among perplexity (Vp) on word-unit (or phoneme-unit), sentence length (L), word (or phoneme) recognition rate (Rw) and sentence recognition rate. So, from this relationship, we can predict the sentence recognition rate, if the word (or phoneme) recognition performance and task defintition are given. The approximation equation is follows: Sentence recognition rate = {f(Vp,Rw)}L, where f(Vp,Rw) denotes the word recognition rate for the vocabulary size Vp obtained by using this recognizer (Rw ).
Cite as: Nakagawa, S. (1989) An evaluation method for continuous speech recognition systems. Proc. Speech Input/Output Assessment and Speech Databases, Vol.2, 151-154
@inproceedings{nakagawa89_sioa,
author={Seiichi Nakagawa},
title={{An evaluation method for continuous speech recognition systems}},
year=1989,
booktitle={Proc. Speech Input/Output Assessment and Speech Databases},
pages={Vol.2, 151-154}
}