Wav2vec 2.0 Fine-Tuning Performance with Expanded Flemish Dutch Pre-Training Data

Recent research in speech processing exhibits a growing interest in unsupervised and self-supervised representation learning from unlabelled data to alleviate the need for large amounts of annotated data. We investigate several popular pre-training methods and apply them to Flemish Dutch. We compare off-the-shelf English pre-trained models to models trained on an increasing amount of Flemish data. We find that the most important factors for positive transfer to downstream speech recognition tasks include a substantial amount of data and a matching pre-training domain. Ideally, we also finetune

Research goal: What is the impact of increasing the pre-training dataset size for Flemish Dutch on the WER (Word Error Rate) of fine-tuned wav2vec 2.0 compared to multilingual pre-trained models like XLS-R?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 7.5/10.