In this paper we present a simple speech recognition application in which an explicit knowledge base system is coupled to a connectionist machine. A set of Prolog rules describes the structural aspects of the words in the vocabulary, in terms of acoustico-phonetic events. These events are jointly evaluated by a multi-layer network in order to identify the corresponding word. The system provides good results, notably with unknown speakers.
Cite as: Bulot, R., Nocera, P. (1989) Explicit knowledge and neural networks for speech recognition. Proc. First European Conference on Speech Communication and Technology (Eurospeech 1989), 2533-2536, doi: 10.21437/Eurospeech.1989-90
@inproceedings{bulot89_eurospeech,
author={R. Bulot and P. Nocera},
title={{Explicit knowledge and neural networks for speech recognition}},
year=1989,
booktitle={Proc. First European Conference on Speech Communication and Technology (Eurospeech 1989)},
pages={2533--2536},
doi={10.21437/Eurospeech.1989-90}
}