Enforcing Encoder-Decoder Modularity in Sequence-to-Sequence Models
arXiv:1911.03782
Abstract
Inspired by modular software design principles of independence, interchangeability, and clarity of interface, we introduce a method for enforcing encoder-decoder modularity in seq2seq models without sacrificing the overall model quality or its full differentiability. We discretize the encoder output units into a predefined interpretable vocabulary space using the Connectionist Temporal Classification (CTC) loss. Our modular systems achieve near SOTA performance on the 300h Switchboard benchmark, with WER of 8.3% and 17.6% on the SWB and CH subsets, using seq2seq models with encoder and decoder modules which are independent and interchangeable.
References in corpus (10)
- SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition
- The information bottleneck method
- A Comparative Study on Transformer vs RNN in Speech Applications
- Disentangling factors of variation in deep representations using adversarial training
- Modular Multitask Reinforcement Learning with Policy Sketches
- fairseq: A Fast, Extensible Toolkit for Sequence Modeling
- Transformers with convolutional context for ASR
- Deep Learning of Representations: Looking Forward
- On the Choice of Modeling Unit for Sequence-to-Sequence Speech Recognition
- A Fully Differentiable Beam Search Decoder