Data Efficient Direct Speech-to-Text Translation with Modality Agnostic Meta-Learning

TL;DR

This paper introduces a modality agnostic meta-learning approach for end-to-end speech translation that leverages transfer learning from ASR and MT tasks, significantly improving translation quality especially in low-data scenarios.

Contribution

The paper proposes a novel meta-learning framework that trains a multi-task model to transfer knowledge from ASR and MT to speech translation, outperforming previous transfer learning methods.

Findings

Achieved state-of-the-art BLEU scores on En-De and En-Fr translation tasks.

Outperformed previous transfer learning approaches by large margins.

Demonstrated effectiveness in low-resource speech translation scenarios.

Abstract

End-to-end Speech Translation (ST) models have several advantages such as lower latency, smaller model size, and less error compounding over conventional pipelines that combine Automatic Speech Recognition (ASR) and text Machine Translation (MT) models. However, collecting large amounts of parallel data for ST task is more difficult compared to the ASR and MT tasks. Previous studies have proposed the use of transfer learning approaches to overcome the above difficulty. These approaches benefit from weakly supervised training data, such as ASR speech-to-transcript or MT text-to-text translation pairs. However, the parameters in these models are updated independently of each task, which may lead to sub-optimal solutions. In this work, we adopt a meta-learning algorithm to train a modality agnostic multi-task model that transfers knowledge from source tasks=ASR+MT to target task=ST where ST task severely lacks data. In the meta-learning phase, the parameters of the model are exposed to vast amounts of speech transcripts (e.g., English ASR) and text translations (e.g., English-German MT). During this phase, parameters are updated in such a way to understand speech, text representations, the relation between them, as well as act as a good initialization point for the target ST task. We evaluate the proposed meta-learning approach for ST tasks on English-German (En-De) and English-French (En-Fr) language pairs from the Multilingual Speech Translation Corpus (MuST-C). Our method outperforms the previous transfer learning approaches and sets new state-of-the-art results for En-De and En-Fr ST tasks by obtaining 9.18, and 11.76 BLEU point improvements, respectively.