Integrating Pretrained ASR and LM to Perform Sequence Generation for Spoken Language Understanding

TL;DR

This paper introduces a three-pass end-to-end SLU system that effectively combines pretrained ASR and LM models for improved sequence generation, especially on challenging acoustic data.

Contribution

It proposes a novel three-pass architecture that integrates ASR and LM subnetworks into a unified SLU framework for better performance.

Findings

Outperforms cascaded and E2E SLU models on SLURP and SLUE datasets.

Shows significant improvements on acoustically challenging utterances.

Demonstrates effective integration of pretrained ASR and LM models.

Abstract

There has been an increased interest in the integration of pretrained speech recognition (ASR) and language models (LM) into the SLU framework. However, prior methods often struggle with a vocabulary mismatch between pretrained models, and LM cannot be directly utilized as they diverge from its NLU formulation. In this study, we propose a three-pass end-to-end (E2E) SLU system that effectively integrates ASR and LM subnetworks into the SLU formulation for sequence generation tasks. In the first pass, our architecture predicts ASR transcripts using the ASR subnetwork. This is followed by the LM subnetwork, which makes an initial SLU prediction. Finally, in the third pass, the deliberation subnetwork conditions on representations from the ASR and LM subnetworks to make the final prediction. Our proposed three-pass SLU system shows improved performance over cascaded and E2E SLU models on two benchmark SLU datasets, SLURP and SLUE, especially on acoustically challenging utterances.