How to adapt Kaldi models? (in Russian)

"How to train a Russian-language model to recognise gamers' speech?" Such questions are asked by those who are interested in and involved in NLP. In particular, NLP specialists are interested in how to adapt the Kaldi models to their domain in order to improve the quality of measurement. This is what we will analyse in this article.

Overview of speech enhancement and noise reduction methods: from classics to SotA (in Russian)

Team at MTS AI has experimented with different types of noise reduction as a pre-processing step for speech recognition. As we will discuss below, this application of noise reduction is somewhat different from noise reduction used to improve the sound of audio to the human ear.