Exploring neural transducers for end-to-end speech recognition

TLDR

The study empirically compares CTC, RNN‑Transducer, and attention‑based Seq2Seq models for end‑to‑end speech recognition. The authors evaluate these models using different encoder architectures, including forward‑only layers and aggressive downsampling, on benchmark and internal datasets. Seq2Seq and RNN‑Transducer models outperform CTC on Hub5'00 without a language model, and RNN‑Transducer with language‑model rescoring remains superior on an internal dataset, simplifying the pipeline to neural‑network‑only decoding.

Abstract

In this work, we perform an empirical comparison among the CTC, RNN-Transducer, and attention-based Seq2Seq models for end-to-end speech recognition. We show that, without any language model, Seq2Seq and RNN-Transducer models both outperform the best reported CTC models with a language model, on the popular Hub5'00 benchmark. On our internal diverse dataset, these trends continue — RNN-Transducer models rescored with a language model after beam search outperform our best CTC models. These results simplify the speech recognition pipeline so that decoding can now be expressed purely as neural network operations. We also study how the choice of encoder architecture affects the performance of the three models — when all encoder layers are forward only, and when encoders downsample the input representation aggressively.