A New Training Pipeline for an Improved Neural Transducer

Abstract

The RNN transducer is a promising end-to-end model candidate. We compare the original training criterion with the full marginalization over all alignments, to the commonly used maximum approximation, which simplifies, improves and speeds up our training. We also generalize from the original neural network model and study more powerful models, made possible due to the maximum approximation. We further generalize the output label topology to cover RNN-T, RNA and CTC. We perform several studies among all these aspects, including a study on the effect of external alignments. We find that the transducer model generalizes much better on longer sequences than the attention model. Our final transducer model outperforms our attention model on Switchboard 300h by over 6% relative WER.

arXiv commentpublished at Interspeech 2020

References in corpus (16)

• SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition
• A Structured Self-attentive Sentence Embedding
• Focal Loss for Dense Object Detection
• Sequence Transduction with Recurrent Neural Networks
• A Comparative Study on Transformer vs RNN in Speech Applications
• Improved training of end-to-end attention models for speech recognition
• A Comprehensive Study of Deep Bidirectional LSTM RNNs for Acoustic Modeling in Speech Recognition
• RETURNN as a Generic Flexible Neural Toolkit with Application to Translation and Speech Recognition
• ContextNet: Improving Convolutional Neural Networks for Automatic Speech Recognition with Global Context
• Transformer-Transducer: End-to-End Speech Recognition with Self-Attention
• Fast-Slow Recurrent Neural Networks
• RNN-T For Latency Controlled ASR With Improved Beam Search
• Monotonic Chunkwise Attention
• Towards a Competitive End-to-End Speech Recognition for CHiME-6 Dinner Party Transcription
• RNN-T Models Fail to Generalize to Out-of-Domain Audio: Causes and Solutions
• An investigation of phone-based subword units for end-to-end speech recognition

Cited by in corpus (10)

• A CTC Alignment-based Non-autoregressive Transformer for End-to-end Automatic Speech Recognition
• Efficient Training of Neural Transducer for Speech Recognition
• Advanced Long-Content Speech Recognition With Factorized Neural Transducer
• A study of latent monotonic attention variants
• A systematic comparison of grapheme-based vs. phoneme-based label units for encoder-decoder-attention models
• Equivalence of Segmental and Neural Transducer Modeling: A Proof of Concept
• On lattice-free boosted MMI training of HMM and CTC-based full-context ASR models
• A Light-weight contextual spelling correction model for customizing transducer-based speech recognition systems
• Phoneme Based Neural Transducer for Large Vocabulary Speech Recognition
• CASS-NAT: CTC Alignment-based Single Step Non-autoregressive Transformer for Speech Recognition