HebDB: a Weakly Supervised Dataset for Hebrew Speech Processing

TL;DR

HebDB is a new large-scale, weakly supervised Hebrew speech dataset that enables improved speech recognition tools, demonstrated by baseline models outperforming multilingual alternatives.

Contribution

The paper introduces HebDB, a substantial Hebrew speech dataset with baseline ASR models, advancing Hebrew speech processing research.

Findings

Baseline models outperform multilingual ASR on HebDB

HebDB contains 2500 hours of natural Hebrew speech

Models trained on HebDB show improved performance

Abstract

We present HebDB, a weakly supervised dataset for spoken language processing in the Hebrew language. HebDB offers roughly 2500 hours of natural and spontaneous speech recordings in the Hebrew language, consisting of a large variety of speakers and topics. We provide raw recordings together with a pre-processed, weakly supervised, and filtered version. The goal of HebDB is to further enhance research and development of spoken language processing tools for the Hebrew language. Hence, we additionally provide two baseline systems for Automatic Speech Recognition (ASR): (i) a self-supervised model; and (ii) a fully supervised model. We present the performance of these two methods optimized on HebDB and compare them to current multi-lingual ASR alternatives. Results suggest the proposed method reaches better results than the evaluated baselines considering similar model sizes. Dataset, code, and models are publicly available under https://pages.cs.huji.ac.il/adiyoss-lab/HebDB/.