Segmental Recurrent Neural Networks for End-to-end Speech Recognition

TL;DR

This paper introduces a self-contained, end-to-end segmental RNN-CRF model for speech recognition that jointly learns features and segmentation, achieving state-of-the-art results on TIMIT without external features or language models.

Contribution

It presents a novel end-to-end segmental RNN-CRF model that marginalizes segmentation, removing the need for external systems and enabling joint training for speech recognition.

Findings

Achieved 17.3% PER on TIMIT, the best for CRF-based models.

Model does not rely on external features or language models.

Demonstrated practical training and decoding methods for speech recognition.

Abstract

We study the segmental recurrent neural network for end-to-end acoustic modelling. This model connects the segmental conditional random field (CRF) with a recurrent neural network (RNN) used for feature extraction. Compared to most previous CRF-based acoustic models, it does not rely on an external system to provide features or segmentation boundaries. Instead, this model marginalises out all the possible segmentations, and features are extracted from the RNN trained together with the segmental CRF. In essence, this model is self-contained and can be trained end-to-end. In this paper, we discuss practical training and decoding issues as well as the method to speed up the training in the context of speech recognition. We performed experiments on the TIMIT dataset. We achieved 17.3 phone error rate (PER) from the first-pass decoding --- the best reported result using CRFs, despite the fact that we only used a zeroth-order CRF and without using any language model.