End-to-End Speech Translation for Low-Resource Languages Using Weakly Labeled Data

TL;DR

This paper demonstrates that weakly labeled data can effectively be used to develop end-to-end speech translation systems for low-resource languages, achieving performance comparable to large-scale baselines.

Contribution

It introduces a novel approach to leverage weakly labeled data and constructs new multilingual speech translation datasets for low-resource languages.

Findings

Weakly labeled data can produce competitive speech translation models.

Constructed the Shrutilipi-anuvaad dataset for multiple low-resource language pairs.

Performance is comparable to large multilingual baselines.

Abstract

The scarcity of high-quality annotated data presents a significant challenge in developing effective end-to-end speech-to-text translation (ST) systems, particularly for low-resource languages. This paper explores the hypothesis that weakly labeled data can be used to build ST models for low-resource language pairs. We constructed speech-to-text translation datasets with the help of bitext mining using state-of-the-art sentence encoders. We mined the multilingual Shrutilipi corpus to build Shrutilipi-anuvaad, a dataset comprising ST data for language pairs Bengali-Hindi, Malayalam-Hindi, Odia-Hindi, and Telugu-Hindi. We created multiple versions of training data with varying degrees of quality and quantity to investigate the effect of quality versus quantity of weakly labeled data on ST model performance. Results demonstrate that ST systems can be built using weakly labeled data, with performance comparable to massive multi-modal multilingual baselines such as SONAR and SeamlessM4T.