Transfer Learning from Pre-trained Language Models Improves End-to-End Speech Summarization

TL;DR

This paper introduces a transfer learning approach that integrates pre-trained language models into end-to-end speech summarization systems, significantly improving naturalness and performance despite limited training data.

Contribution

It is the first to incorporate pre-trained language models into E2E speech summarization and transfers the encoder to bridge the gap between pre-trained components.

Findings

Outperforms baseline models in speech summarization tasks.

Enhances naturalness of generated summaries.

Effective even with limited training data.

Abstract

End-to-end speech summarization (E2E SSum) directly summarizes input speech into easy-to-read short sentences with a single model. This approach is promising because it, in contrast to the conventional cascade approach, can utilize full acoustical information and mitigate to the propagation of transcription errors. However, due to the high cost of collecting speech-summary pairs, an E2E SSum model tends to suffer from training data scarcity and output unnatural sentences. To overcome this drawback, we propose for the first time to integrate a pre-trained language model (LM), which is highly capable of generating natural sentences, into the E2E SSum decoder via transfer learning. In addition, to reduce the gap between the independently pre-trained encoder and decoder, we also propose to transfer the baseline E2E SSum encoder instead of the commonly used automatic speech recognition encoder. Experimental results show that the proposed model outperforms baseline and data augmented models.