Hard-mining for ASR

Student: Ivan Ershov

Educational Programme: Applied Mathematics and Information Science (Bachelor)

Year of Graduation: 2024

It is a powerful way to enhance Automatic Speech Recognition (ASR) systems perfomance by training the model (from scratch or further trained) with a larger amount of higher quality data. However, collecting quality data for ASR training might be challenging, as the concentration of complex examples in a large amount of data is very low. We introduce hard samples mining method. It uses a complexity classifier to predict the probability of the audio recording to be a hard one, so that we could filter these recordings from a huge unlabeled dataset and then train main ASR system on the filtered data. Thus, we use less resources still maintaining or improving model scores.

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