We explore unifying a neural segmenter with two-pass cascaded encoder ASR into a single model. A key challenge is allowing the segmenter (which runs in real-time, synchronously with the decoder) to finalize the 2nd pass (which runs 900 ms behind real-time) without introducing user-perceived latency or deletion errors during inference. We propose a design where the neural segmenter is integrated with the causal 1st pass decoder to emit a end-of-segment (EOS) signal in real-time. The EOS signal is then used to finalize the non-causal 2nd pass. We experiment with different ways to finalize the 2nd pass, and find that a novel dummy frame injection strategy allows for simultaneous high quality 2nd pass results and low finalization latency. On a real-world long-form captioning task (YouTube), we achieve 2.4% relative WER and 140 ms EOS latency gains over a baseline VAD-based segmenter with the same cascaded encoder.
@article{arxiv.2211.15432,
title = {E2E Segmentation in a Two-Pass Cascaded Encoder ASR Model},
author = {W. Ronny Huang and Shuo-Yiin Chang and Tara N. Sainath and Yanzhang He and David Rybach and Robert David and Rohit Prabhavalkar and Cyril Allauzen and Cal Peyser and Trevor D. Strohman},
journal= {arXiv preprint arXiv:2211.15432},
year = {2023}
}