N-Best ASR Transformer: Enhancing SLU Performance using Multiple ASR Hypotheses

TL;DR

This paper proposes using concatenated N-best ASR hypotheses as input to transformer models for SLU, improving performance especially in low-data scenarios and enabling use with third-party ASR APIs.

Contribution

It introduces a simple yet effective method of leveraging multiple ASR hypotheses with transformer models for enhanced SLU performance, especially under limited data conditions.

Findings

Achieved state-of-the-art results on DSTC2 dataset.

Significantly outperformed prior methods in low-data regimes.

Method is compatible with third-party ASR APIs lacking lattice info.

Abstract

Spoken Language Understanding (SLU) systems parse speech into semantic structures like dialog acts and slots. This involves the use of an Automatic Speech Recognizer (ASR) to transcribe speech into multiple text alternatives (hypotheses). Transcription errors, common in ASRs, impact downstream SLU performance negatively. Approaches to mitigate such errors involve using richer information from the ASR, either in form of N-best hypotheses or word-lattices. We hypothesize that transformer models learn better with a simpler utterance representation using the concatenation of the N-best ASR alternatives, where each alternative is separated by a special delimiter [SEP]. In our work, we test our hypothesis by using concatenated N-best ASR alternatives as the input to transformer encoder models, namely BERT and XLM-RoBERTa, and achieve performance equivalent to the prior state-of-the-art model on DSTC2 dataset. We also show that our approach significantly outperforms the prior state-of-the-art when subjected to the low data regime. Additionally, this methodology is accessible to users of third-party ASR APIs which do not provide word-lattice information.