Pseudo-Labeling for Domain-Agnostic Bangla Automatic Speech Recognition
This addresses the problem of developing ASR for low-resource languages like Bangla by creating a domain-agnostic dataset, though it is incremental as it applies pseudo-labeling, an existing method, to a new language context.
The study tackled the challenge of limited labeled data for low-resource languages in automatic speech recognition by proposing a pseudo-labeling approach to create a large-scale domain-agnostic Bangla speech dataset of over 20,000 hours, and used it to train a conformer-based ASR system that demonstrated efficacy on a human-annotated test set and public datasets.
One of the major challenges for developing automatic speech recognition (ASR) for low-resource languages is the limited access to labeled data with domain-specific variations. In this study, we propose a pseudo-labeling approach to develop a large-scale domain-agnostic ASR dataset. With the proposed methodology, we developed a 20k+ hours labeled Bangla speech dataset covering diverse topics, speaking styles, dialects, noisy environments, and conversational scenarios. We then exploited the developed corpus to design a conformer-based ASR system. We benchmarked the trained ASR with publicly available datasets and compared it with other available models. To investigate the efficacy, we designed and developed a human-annotated domain-agnostic test set composed of news, telephony, and conversational data among others. Our results demonstrate the efficacy of the model trained on psuedo-label data for the designed test-set along with publicly-available Bangla datasets. The experimental resources will be publicly available.(https://github.com/hishab-nlp/Pseudo-Labeling-for-Domain-Agnostic-Bangla-ASR)