Investigating Self-supervised, Weakly Supervised And Fully Supervised Training Approaches For Multi-domain Automatic Speech Recognition: A Study On Bangladeshi Bangla

Abstract

Despite huge improvements in automatic speech recognition (ASR) employing neural networks, ASR systems still suffer from a lack of robustness and generalizability issues due to domain shifting. This is mainly because principal corpus design criteria are often not identified and examined adequately while compiling ASR datasets. In this study, we investigate the robustness of the state-of-the-art transfer learning approaches such as self-supervised wav2vec 2.0 and weakly supervised Whisper as well as fully supervised convolutional neural networks (CNNs) for multi-domain ASR. We also demonstrate the significance of domain selection while building a corpus by assessing these models on a novel multi-domain Bangladeshi Bangla ASR evaluation benchmark - BanSpeech, which contains approximately 6.52 hours of human-annotated speech and 8085 utterances from 13 distinct domains. SUBAK.KO, a mostly read speech corpus for the morphologically rich language Bangla, has been used to train the ASR syste

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keysamin2022investigating

Related papers

• Bangla-wave: Improving Bangla Automatic Speech Recognition Utilizing N-gram Language Models (2022)5.24
• An Automatic Speech Recognition System For Bengali Language Based On Wav2vec2 And Transfer Learning (2022)0.00
• Analyzing The Robustness Of Unsupervised Speech Recognition (2021)7.81
• Accent-robust Automatic Speech Recognition Using Supervised And Unsupervised Wav2vec Embeddings (2021)0.00
• Multi-level Embedding Conformer Framework For Bengali Automatic Speech Recognition (2025)0.00
• Boosting Cross-domain Speech Recognition With Self-supervision (2022)0.00
• Toward Domain-invariant Speech Recognition Via Large Scale Training (2018)13.39
• Analyzing The Factors Affecting Usefulness Of Self-supervised Pre-trained Representations For Speech Recognition (2022)0.00