A data generation pipeline that converts sentence-level Swiss German recordings into long-form audio lets a fine-tuned Whisper Large-v2 beat prior state-of-the-art transcription models without losing timestamp prediction.
Keyword-Guided Adaptation of Automatic Speech Recognition
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
Automatic Speech Recognition (ASR) technology has made significant progress in recent years, providing accurate transcription across various domains. However, some challenges remain, especially in noisy environments and specialized jargon. In this paper, we propose a novel approach for improved jargon word recognition by contextual biasing Whisper-based models. We employ a keyword spotting model that leverages the Whisper encoder representation to dynamically generate prompts for guiding the decoder during the transcription process. We introduce two approaches to effectively steer the decoder towards these prompts: KG-Whisper, which is aimed at fine-tuning the Whisper decoder, and KG-Whisper-PT, which learns a prompt prefix. Our results show a significant improvement in the recognition accuracy of specified keywords and in reducing the overall word error rates. Specifically, in unseen language generalization, we demonstrate an average WER improvement of 5.1% over Whisper.
fields
cs.CL 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Fine-tuning Whisper on Low-Resource Languages for Real-World Applications
A data generation pipeline that converts sentence-level Swiss German recordings into long-form audio lets a fine-tuned Whisper Large-v2 beat prior state-of-the-art transcription models without losing timestamp prediction.