Performance Comparison of Pre-trained Models for Speech-to-Text in Turkish: Whisper-Small and Wav2Vec2-XLS-R-300M

TL;DR

This study compares the speech-to-text performance of Whisper-Small and Wav2Vec2-XLS-R-300M models on Turkish, using open-source data and real-world call center recordings, highlighting their accuracy differences.

Contribution

It provides a comparative analysis of two pre-trained multilingual speech-to-text models specifically for Turkish, including fine-tuning and evaluation on real-world data.

Findings

Wav2Vec2-XLS-R-300M achieved a WER of 0.28

Whisper-Small achieved a WER of 0.16

Models' performances were validated on unseen call center data

Abstract

In this study, the performances of the Whisper-Small and Wav2Vec2-XLS-R-300M models which are two pre-trained multilingual models for speech to text were examined for the Turkish language. Mozilla Common Voice version 11.0 which is prepared in Turkish language and is an open-source data set, was used in the study. The multilingual models, Whisper- Small and Wav2Vec2-XLS-R-300M were fine-tuned with this data set which contains a small amount of data. The speech to text performance of the two models was compared. WER values are calculated as 0.28 and 0.16 for the Wav2Vec2-XLS- R-300M and the Whisper-Small models respectively. In addition, the performances of the models were examined with the test data prepared with call center records that were not included in the training and validation dataset.