2M-BELEBELE: Highly Multilingual Speech and American Sign Language Comprehension Dataset

TL;DR

This paper presents 2M-BELEBELE, a comprehensive multilingual speech and ASL dataset extending BELEBELE, enabling evaluation of cross-lingual comprehension with promising accuracy results.

Contribution

It introduces the first highly multilingual speech and ASL dataset, covering 74 spoken languages and one sign language, expanding resources for multilingual comprehension research.

Findings

Speech comprehension accuracy is ~2-3% lower than reading comprehension across languages.

The dataset enables evaluation in 5-shot and zero-shot learning settings.

It demonstrates the feasibility of multilingual speech and sign language understanding.

Abstract

We introduce the first highly multilingual speech and American Sign Language (ASL) comprehension dataset by extending BELEBELE. Our dataset covers 74 spoken languages at the intersection of BELEBELE and FLEURS, and one sign language (ASL). We evaluate 2M-BELEBELE dataset for both 5-shot and zero-shot settings and across languages, the speech comprehension accuracy is ~ 2-3% average lower compared to reading comprehension.