Search papers, labs, and topics across Lattice.
This study compares context biasing methods and speech large language models (LLMs) for recognizing new and rare words in automatic speech recognition (ASR). The context biasing methods significantly reduced biased word error rates (WER) by up to 88% without adversely affecting the recognition of other words, while speech LLMs demonstrated strong performance on read speech but struggled with non-read speech due to sensitivity to distractor count and prompt word order. The findings highlight the trade-offs between these approaches, providing valuable insights for selecting the most effective method for different speech contexts.
Context biasing methods can slash biased word error rates by up to 88%, outperforming speech LLMs in challenging ASR scenarios.
Recognizing new and rare words - named entities, acronyms, domain specific special words, and other items scarce in training data - remains a key challenge for automatic speech recognition (ASR). We compare two strategies for this: context biasing methods, where an ASR model is extended such that during inference a word list can be supplied, and speech large language models (LLMs) prompted with context directly. We evaluate two context biasing methods based on Whisper against three speech LLMs across read and non-read speech, reporting biased, unbiased, and overall word error rate (WER). The context biasing methods cut biased WER by up to 88% relative while leaving other words largely unaffected. Speech LLMs excel on read speech but generalize less well to non-read speech, and prove sensitive to distractor count and prompt word order. We characterize the resulting trade-offs to guide method selection.
