A Comparative Study on End-to-end Speech to Text Translation
This work provides incremental improvements for speech translation researchers by optimizing model performance on specific datasets.
The study compared end-to-end speech-to-text translation architectures and pre-training strategies, achieving up to 4% BLEU and 5% TER improvements and outperforming state-of-the-art systems on IWSLT and LibriSpeech datasets.
Recent advances in deep learning show that end-to-end speech to text translation model is a promising approach to direct the speech translation field. In this work, we provide an overview of different end-to-end architectures, as well as the usage of an auxiliary connectionist temporal classification (CTC) loss for better convergence. We also investigate on pre-training variants such as initializing different components of a model using pre-trained models, and their impact on the final performance, which gives boosts up to 4% in BLEU and 5% in TER. Our experiments are performed on 270h IWSLT TED-talks En->De, and 100h LibriSpeech Audiobooks En->Fr. We also show improvements over the current end-to-end state-of-the-art systems on both tasks.