In the paper, we describe a research of factored language model (FLM) for Russian speech recognition. We used FLM at N-best list rescoring stage. Optimization of the FLM parameters was carried out by means of Genetic Algorithm. The best models used four factors: lemma, morphological tag, stem, and word. Experiments on large vocabulary continuous Russian speech recognition showed a relative WER reduction of 8% when FLM was interpolated with the baseline 3-gram model.
CITATION STYLE
Kipyatkova, I., & Karpov, A. (2014). Study of morphological factors of factored language models for russian ASR. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8773, pp. 451–458). Springer Verlag. https://doi.org/10.1007/978-3-319-11581-8_56
Mendeley helps you to discover research relevant for your work.