LT-LM: a novel non-autoregressive language model for single-shot lattice rescoring

TL;DR

The paper introduces LT-LM, a non-autoregressive lattice transformer language model that enables single-shot rescoring of speech recognition lattices, significantly increasing speed with minimal impact on accuracy.

Contribution

It presents a novel non-autoregressive lattice transformer model and an artificial lattice generation method for efficient single-pass rescoring in speech recognition.

Findings

Single-shot rescoring is over 300 times faster than traditional methods.

LT-LM achieves comparable WER with significantly reduced computational cost.

Artificial lattice generation enhances training data for the model.

Abstract

Neural network-based language models are commonly used in rescoring approaches to improve the quality of modern automatic speech recognition (ASR) systems. Most of the existing methods are computationally expensive since they use autoregressive language models. We propose a novel rescoring approach, which processes the entire lattice in a single call to the model. The key feature of our rescoring policy is a novel non-autoregressive Lattice Transformer Language Model (LT-LM). This model takes the whole lattice as an input and predicts a new language score for each arc. Additionally, we propose the artificial lattices generation approach to incorporate a large amount of text data in the LT-LM training process. Our single-shot rescoring performs orders of magnitude faster than other rescoring methods in our experiments. It is more than 300 times faster than pruned RNNLM lattice rescoring and N-best rescoring while slightly inferior in terms of WER.