SpeechBERT: An Audio-and-text Jointly Learned Language Model for End-to-end Spoken Question Answering

TL;DR

SpeechBERT is a novel end-to-end audio-and-text model for spoken question answering that outperforms traditional cascade methods, demonstrating robustness to ASR errors and potential for broader spoken language understanding tasks.

Contribution

This paper introduces SpeechBERT, the first end-to-end model for spoken question answering that jointly learns from audio and text, outperforming cascade approaches.

Findings

End-to-end SpeechBERT outperforms cascade models on datasets with ASR errors.

Ensembling SpeechBERT with cascade models yields even better results.

SpeechBERT shows potential for various spoken language understanding tasks.

Abstract

While various end-to-end models for spoken language understanding tasks have been explored recently, this paper is probably the first known attempt to challenge the very difficult task of end-to-end spoken question answering (SQA). Learning from the very successful BERT model for various text processing tasks, here we proposed an audio-and-text jointly learned SpeechBERT model. This model outperformed the conventional approach of cascading ASR with the following text question answering (TQA) model on datasets including ASR errors in answer spans, because the end-to-end model was shown to be able to extract information out of audio data before ASR produced errors. When ensembling the proposed end-to-end model with the cascade architecture, even better performance was achieved. In addition to the potential of end-to-end SQA, the SpeechBERT can also be considered for many other spoken language understanding tasks just as BERT for many text processing tasks.