Zero-shot Context Biasing with Trie-based Decoding using Synthetic Multi-Pronunciation
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912565677785088 |
|---|---|
| author | Liu, Changsong Peng, Yizhou Chng, Eng Siong |
| author_facet | Liu, Changsong Peng, Yizhou Chng, Eng Siong |
| contents | Contextual automatic speech recognition (ASR) systems allow for recognizing out-of-vocabulary (OOV) words, such as named entities or rare words. However, it remains challenging due to limited training data and ambiguous or inconsistent pronunciations. In this paper, we propose a synthesis-driven multi-pronunciation contextual biasing method that performs zero-shot contextual ASR on a pretrained Whisper model. Specifically, we leverage text-to-speech (TTS) systems to synthesize diverse speech samples containing each target rare word, and then use the pretrained Whisper model to extract multiple predicted pronunciation variants. These variant token sequences are compiled into a prefix-trie, which assigns rewards to beam hypotheses in a shallow-fusion manner during beam-search decoding. Subsequently, any recognized variant is mapped back to the original rare word in the final transcription. The evaluation results on the LibriSpeech dataset show that our method reduces biased-word error rate (B-WER) by 43% on test-clean and 44% on test-other while maintaining unbiased-WER (U-WER) essentially unchanged. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_17796 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Zero-shot Context Biasing with Trie-based Decoding using Synthetic Multi-Pronunciation Liu, Changsong Peng, Yizhou Chng, Eng Siong Computation and Language Audio and Speech Processing Contextual automatic speech recognition (ASR) systems allow for recognizing out-of-vocabulary (OOV) words, such as named entities or rare words. However, it remains challenging due to limited training data and ambiguous or inconsistent pronunciations. In this paper, we propose a synthesis-driven multi-pronunciation contextual biasing method that performs zero-shot contextual ASR on a pretrained Whisper model. Specifically, we leverage text-to-speech (TTS) systems to synthesize diverse speech samples containing each target rare word, and then use the pretrained Whisper model to extract multiple predicted pronunciation variants. These variant token sequences are compiled into a prefix-trie, which assigns rewards to beam hypotheses in a shallow-fusion manner during beam-search decoding. Subsequently, any recognized variant is mapped back to the original rare word in the final transcription. The evaluation results on the LibriSpeech dataset show that our method reduces biased-word error rate (B-WER) by 43% on test-clean and 44% on test-other while maintaining unbiased-WER (U-WER) essentially unchanged. |
| title | Zero-shot Context Biasing with Trie-based Decoding using Synthetic Multi-Pronunciation |
| topic | Computation and Language Audio and Speech Processing |
| url | https://arxiv.org/abs/2508.17796 |