LLM-based Generative Error Correction for Rare Words with Synthetic Data and Phonetic Context
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918031927541760 |
|---|---|
| author | Yamashita, Natsuo Yamamoto, Masaaki Kokubo, Hiroaki Kawaguchi, Yohei |
| author_facet | Yamashita, Natsuo Yamamoto, Masaaki Kokubo, Hiroaki Kawaguchi, Yohei |
| contents | Generative error correction (GER) with large language models (LLMs) has emerged as an effective post-processing approach to improve automatic speech recognition (ASR) performance. However, it often struggles with rare or domain-specific words due to limited training data. Furthermore, existing LLM-based GER approaches primarily rely on textual information, neglecting phonetic cues, which leads to over-correction. To address these issues, we propose a novel LLM-based GER approach that targets rare words and incorporates phonetic information. First, we generate synthetic data to contain rare words for fine-tuning the GER model. Second, we integrate ASR's N-best hypotheses along with phonetic context to mitigate over-correction. Experimental results show that our method not only improves the correction of rare words but also reduces the WER and CER across both English and Japanese datasets. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_17410 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | LLM-based Generative Error Correction for Rare Words with Synthetic Data and Phonetic Context Yamashita, Natsuo Yamamoto, Masaaki Kokubo, Hiroaki Kawaguchi, Yohei Sound Computation and Language Audio and Speech Processing Generative error correction (GER) with large language models (LLMs) has emerged as an effective post-processing approach to improve automatic speech recognition (ASR) performance. However, it often struggles with rare or domain-specific words due to limited training data. Furthermore, existing LLM-based GER approaches primarily rely on textual information, neglecting phonetic cues, which leads to over-correction. To address these issues, we propose a novel LLM-based GER approach that targets rare words and incorporates phonetic information. First, we generate synthetic data to contain rare words for fine-tuning the GER model. Second, we integrate ASR's N-best hypotheses along with phonetic context to mitigate over-correction. Experimental results show that our method not only improves the correction of rare words but also reduces the WER and CER across both English and Japanese datasets. |
| title | LLM-based Generative Error Correction for Rare Words with Synthetic Data and Phonetic Context |
| topic | Sound Computation and Language Audio and Speech Processing |
| url | https://arxiv.org/abs/2505.17410 |