Whispering LLaMA: A Cross-Modal Generative Error Correction Framework for Speech Recognition

Abstract

We introduce a new cross-modal fusion technique designed for generative error correction in automatic speech recognition (ASR). Our methodology leverages both acoustic information and external linguistic representations to generate accurate speech transcription contexts. This marks a step towards a fresh paradigm in generative error correction within the realm of n-best hypotheses. Unlike the existing ranking-based rescoring methods, our approach adeptly uses distinct initialization techniques and parameter-efficient algorithms to boost ASR performance derived from pre-trained speech and text models. Through evaluation across diverse ASR datasets, we evaluate the stability and reproducibility of our fusion technique, demonstrating its improved word error rate relative (WERR) performance in comparison to n-best hypotheses by relatively 37.66%. To encourage future research, we have made our code and pre-trained models open source at https://github.com/Srijith-rkr/Whispering-LLaMA.

Accepted to EMNLP 2023 as main paper. 10 pages. Revised math notations. GitHub: https://github.com/Srijith-rkr/Whispering-LLaMA

References in corpus (10)

• LLaMA: Open and Efficient Foundation Language Models
• Robust Speech Recognition via Large-Scale Weak Supervision
• Multimodal Chain-of-Thought Reasoning in Language Models
• N-best T5: Robust ASR Error Correction using Multiple Input Hypotheses and Constrained Decoding Space
• From English to More Languages: Parameter-Efficient Model Reprogramming for Cross-Lingual Speech Recognition
• SpeechPrompt v2: Prompt Tuning for Speech Classification Tasks
• A Parameter-Efficient Learning Approach to Arabic Dialect Identification with Pre-Trained General-Purpose Speech Model
• HyPoradise: An Open Baseline for Generative Speech Recognition with Large Language Models
• How to Estimate Model Transferability of Pre-Trained Speech Models?
• Can Whisper perform speech-based in-context learning?

Cited by in corpus (2)

• Large Language Model Based Generative Error Correction: A Challenge and Baselines for Speech Recognition, Speaker Tagging, and Emotion Recognition
• A Multimodal Approach to Device-Directed Speech Detection with Large Language Models