Context-Aware Dynamic Chunking for Streaming Tibetan Speech Recognition

TL;DR

This paper introduces a context-aware dynamic chunking approach for streaming Tibetan speech recognition, improving accuracy and latency by adaptively adjusting chunk sizes and incorporating linguistic and language model information.

Contribution

It presents a novel adaptive chunking mechanism combined with linguistic units and external language models for Tibetan speech recognition.

Findings

Achieved 6.23% WER on test set

48.15% relative improvement over fixed-chunk baseline

Reduced recognition latency significantly

Abstract

In this work, we propose a streaming speech recognition framework for Amdo Tibetan, built upon a hybrid CTC/Atten-tion architecture with a context-aware dynamic chunking mechanism. The proposed strategy adaptively adjusts chunk widths based on encoding states, enabling flexible receptive fields, cross-chunk information exchange, and robust adaptation to varying speaking rates, thereby alleviating the context truncation problem of fixed-chunk methods. To further capture the linguistic characteristics of Tibetan, we construct a lexicon grounded in its orthographic principles, providing linguistically motivated modeling units. During decoding, an external language model is integrated to enhance semantic consistency and improve recognition of long sentences. Experimental results show that the proposed framework achieves a word error rate (WER) of 6.23% on the test set, yielding a 48.15% relative improvement over the fixed-chunk baseline, while significantly reducing recognition latency and maintaining performance close to global decoding.