Research on Speech to Text Transformation Methods

Student: Kilyazov Vladimir

Supervisor: Nikolay Karpov

Faculty: Faculty of Informatics, Mathematics, and Computer Science (HSE Nizhny Novgorod)

Educational Programme: Applied Mathematics and Information Science (Bachelor)

Year of Graduation: 2019

In this paper we focus on development of the Russian language speech recognition system based on DeepSpeech architecture. The system was trained on a custom speech dataset which was collected from YouTube. The language model was developed based on corpus of popular articles in Russian version of Wikipedia. The resulting system was tested on a dataset consisting of audio recordings of Russian literature recorded by more than 25 different speakers, which is known as voxforge.com dataset. The best WER demonstrated by our approach currently equals to 25% with language model and 35% without language model usage.

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses