[go: up one dir, main page]

Fast Streaming Transducer ASR Prototyping via Knowledge Distillation with Whisper

Iuliia Thorbecke, Juan Pablo Zuluaga Gomez, Esaú Villatoro-tello, Shashi Kumar, Pradeep Rangappa, Sergio Burdisso, Petr Motlicek, Karthik Pandia D S, Aravind Ganapathiraju


Abstract
The training of automatic speech recognition (ASR) with little to no supervised data remains an open question. In this work, we demonstrate that streaming Transformer-Transducer (TT) models can be trained from scratch in consumer and accessible GPUs in their entirety with pseudo-labeled (PL) speech from foundational speech models (FSM). This allows training a robust ASR model just in one stage and does not require large data and computational budget compared to the two-step scenario with pre-training and fine-tuning. We perform a comprehensive ablation on different aspects of PL-based streaming TT models such as the impact of (1) shallow fusion of n-gram LMs, (2) contextual biasing with named entities, (3) chunk-wise decoding for low-latency streaming applications, and (4) TT overall performance as the function of the FSM size. Our results demonstrate that TT can be trained from scratch without supervised data, even with very noisy PLs. We validate the proposed framework on 6 languages from CommonVoice and propose multiple heuristics to filter out hallucinated PLs.
Anthology ID:
2024.findings-emnlp.976
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2024
Month:
November
Year:
2024
Address:
Miami, Florida, USA
Editors:
Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
16747–16762
Language:
URL:
https://aclanthology.org/2024.findings-emnlp.976
DOI:
10.18653/v1/2024.findings-emnlp.976
Bibkey:
Cite (ACL):
Iuliia Thorbecke, Juan Pablo Zuluaga Gomez, Esaú Villatoro-tello, Shashi Kumar, Pradeep Rangappa, Sergio Burdisso, Petr Motlicek, Karthik Pandia D S, and Aravind Ganapathiraju. 2024. Fast Streaming Transducer ASR Prototyping via Knowledge Distillation with Whisper. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 16747–16762, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
Fast Streaming Transducer ASR Prototyping via Knowledge Distillation with Whisper (Thorbecke et al., Findings 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.findings-emnlp.976.pdf