End-to-End Speech Recognition Contextualization with Large Language Models
This addresses the challenge of leveraging unstructured contextual information in speech recognition for improved accuracy, particularly on rare words, though it is incremental as it builds on existing LLM capabilities.
The paper tackles the problem of contextualizing speech recognition by incorporating Large Language Models (LLMs) as a mixed-modal language modeling task, resulting in a 6% WER reduction with additional textual context and a 7.5% overall WER improvement over a baseline system.
In recent years, Large Language Models (LLMs) have garnered significant attention from the research community due to their exceptional performance and generalization capabilities. In this paper, we introduce a novel method for contextualizing speech recognition models incorporating LLMs. Our approach casts speech recognition as a mixed-modal language modeling task based on a pretrained LLM. We provide audio features, along with optional text tokens for context, to train the system to complete transcriptions in a decoder-only fashion. As a result, the system is implicitly incentivized to learn how to leverage unstructured contextual information during training. Our empirical results demonstrate a significant improvement in performance, with a 6% WER reduction when additional textual context is provided. Moreover, we find that our method performs competitively and improve by 7.5% WER overall and 17% WER on rare words against a baseline contextualized RNN-T system that has been trained on more than twenty five times larger speech dataset. Overall, we demonstrate that by only adding a handful number of trainable parameters via adapters, we can unlock contextualized speech recognition capability for the pretrained LLM while keeping the same text-only input functionality.