Streaming End-to-End Multilingual Speech Recognition with Joint Language Identification

TL;DR

This paper introduces a streaming multilingual speech recognition model that integrates language identification directly into the RNN-T architecture, achieving high accuracy with minimal additional latency.

Contribution

It proposes a novel RNN-T based model with integrated per-frame language ID predictor for improved streaming multilingual speech recognition.

Findings

Achieves 96.2% language ID accuracy

Maintains same second-pass WER as oracle LID

Operates with low test-time cost

Abstract

Language identification is critical for many downstream tasks in automatic speech recognition (ASR), and is beneficial to integrate into multilingual end-to-end ASR as an additional task. In this paper, we propose to modify the structure of the cascaded-encoder-based recurrent neural network transducer (RNN-T) model by integrating a per-frame language identifier (LID) predictor. RNN-T with cascaded encoders can achieve streaming ASR with low latency using first-pass decoding with no right-context, and achieve lower word error rates (WERs) using second-pass decoding with longer right-context. By leveraging such differences in the right-contexts and a streaming implementation of statistics pooling, the proposed method can achieve accurate streaming LID prediction with little extra test-time cost. Experimental results on a voice search dataset with 9 language locales shows that the proposed method achieves an average of 96.2% LID prediction accuracy and the same second-pass WER as that obtained by including oracle LID in the input.