The aim of the paper is to investigate the ways to improve acoustic models for Russian spontaneous speech recognition. We applied the main steps of the Kaldi Switchboard recipe to a Russian dataset but obtained low accuracy with respect to the results for English spontaneous telephone speech. We found two methods to be especially useful for Russian spontaneous speech: the i-vector based deep neural network adaptation and speaker-dependent bottleneck features which provide 8.6% and 11.9% relative word error rate reduction over the baseline system respectively.
CITATION STYLE
Prudnikov, A., Medennikov, I., Mendelev, V., Korenevsky, M., & Khokhlov, Y. (2015). Improving acoustic models for Russian spontaneous speech recognition. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9319, pp. 234–242). Springer Verlag. https://doi.org/10.1007/978-3-319-23132-7_29
Mendeley helps you to discover research relevant for your work.