E2E Segmentation in a Two-Pass Cascaded Encoder ASR Model

TL;DR

This paper presents a unified neural segmentation approach integrated with a two-pass cascaded encoder ASR model, achieving low latency and high-quality results for long-form captioning.

Contribution

It introduces a novel dummy frame injection strategy enabling real-time segmentation and 2nd pass finalization without latency or errors.

Findings

Achieved 2.4% relative WER reduction on YouTube captioning.

Reduced EOS latency by 140 ms compared to baseline.

Demonstrated effective real-time segmentation in a cascaded ASR system.

Abstract

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.