About this document
Lattice-Based LLMs for ASR Improvement by gogigorgonzola is a document available to read on EtoBox.
This document discusses a study on improving automatic speech recognition (ASR) by replacing n-best hypotheses with lattice structures for generative error correction (GER) using large language models (LLMs). The proposed method aims to reduce hypothesis space collapse and enhance transcription accuracy, particularly in noisy environments. Experimental results demonstrate that using lattice formats leads to improved performance in Japanese speech recognition compared to traditional n-best approaches.
- Author
- gogigorgonzola
- Language
- EN