Phoneme Based Neural Transducer for Large Vocabulary Speech Recognition

TL;DR

This paper introduces a phoneme-based neural transducer model for large vocabulary speech recognition that combines classical and end-to-end methods, achieving competitive results with a simple, efficient approach.

Contribution

It proposes a novel phoneme-based neural transducer with label augmentation, simplified neural structure, and an effective training procedure, advancing speech recognition performance.

Findings

Achieves performance comparable to state-of-the-art on TED-LIUM and Switchboard datasets.

Demonstrates the effectiveness of phoneme label augmentation and simplified neural architecture.

Shows that a phonetic context size of one suffices for optimal results.

Abstract

To join the advantages of classical and end-to-end approaches for speech recognition, we present a simple, novel and competitive approach for phoneme-based neural transducer modeling. Different alignment label topologies are compared and word-end-based phoneme label augmentation is proposed to improve performance. Utilizing the local dependency of phonemes, we adopt a simplified neural network structure and a straightforward integration with the external word-level language model to preserve the consistency of seq-to-seq modeling. We also present a simple, stable and efficient training procedure using frame-wise cross-entropy loss. A phonetic context size of one is shown to be sufficient for the best performance. A simplified scheduled sampling approach is applied for further improvement and different decoding approaches are briefly compared. The overall performance of our best model is comparable to state-of-the-art (SOTA) results for the TED-LIUM Release 2 and Switchboard corpora.