Enforcing Encoder-Decoder Modularity in Sequence-to-Sequence Models
This addresses the need for more modular and interpretable seq2seq models in speech recognition, though it is incremental as it builds on existing methods.
The paper tackled the problem of enforcing encoder-decoder modularity in sequence-to-sequence models by discretizing encoder outputs into an interpretable vocabulary using CTC loss, achieving near state-of-the-art performance with word error rates of 8.3% and 17.6% on SWB and CH subsets of the 300h Switchboard benchmark.
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.