FeruzaSpeech: A 60 Hour Uzbek Read Speech Corpus with Punctuation, Casing, and Context

TL;DR

FeruzaSpeech is a comprehensive 60-hour Uzbek read speech corpus with transcripts in Cyrillic and Latin, designed to improve speech recognition systems and freely accessible for research.

Contribution

The paper introduces FeruzaSpeech, a new high-quality Uzbek speech corpus with diverse content and dual-script transcripts, enhancing speech recognition performance.

Findings

Improved Word Error Rates on Uzbek speech datasets

High-quality recordings from a native speaker

Availability of the corpus for academic research

Abstract

This paper introduces FeruzaSpeech, a read speech corpus of the Uzbek language, containing transcripts in both Cyrillic and Latin alphabets, freely available for academic research purposes. This corpus includes 60 hours of high-quality recordings from a single native female speaker from Tashkent, Uzbekistan. These recordings consist of short excerpts from a book and BBC News. This paper discusses the enhancement of the Word Error Rates (WERs) on CommonVoice 16.1's Uzbek data, Uzbek Speech Corpus data, and FeruzaSpeech data upon integrating FeruzaSpeech.