TED-LIUM 3: twice as much data and corpus repartition for experiments on speaker adaptation

TL;DR

This paper introduces TED-LIUM 3, a significantly larger speech corpus for English ASR, compares recent ASR system developments, and proposes new data partitions for speaker adaptation experiments.

Contribution

It provides a new, larger TED-LIUM 3 corpus with two data repartitions, including one specifically designed for speaker adaptation research.

Findings

End-to-end ASR benefits more from increased data than HMM-based systems.

HMM-based ASR still outperforms end-to-end systems at 452 hours of data.

The corpus and partitions are freely available for research.

Abstract

In this paper, we present TED-LIUM release 3 corpus dedicated to speech recognition in English, that multiplies by more than two the available data to train acoustic models in comparison with TED-LIUM 2. We present the recent development on Automatic Speech Recognition (ASR) systems in comparison with the two previous releases of the TED-LIUM Corpus from 2012 and 2014. We demonstrate that, passing from 207 to 452 hours of transcribed speech training data is really more useful for end-to-end ASR systems than for HMM-based state-of-the-art ones, even if the HMM-based ASR system still outperforms end-to-end ASR system when the size of audio training data is 452 hours, with respectively a Word Error Rate (WER) of 6.6% and 13.7%. Last, we propose two repartitions of the TED-LIUM release 3 corpus: the legacy one that is the same as the one existing in release 2, and a new one, calibrated and designed to make experiments on speaker adaptation. Like the two first releases, TED-LIUM 3 corpus will be freely available for the research community.