Streaming Punctuation: A Novel Punctuation Technique Leveraging Bidirectional Context for Continuous Speech Recognition

While speech recognition Word Error Rate (WER) has reached human parity for English, continuous scenarios such as voice typing and meeting transcriptions still suffer from segmentation punctuation problems, resulting irregular pausing patterns or slow speakers. Transformer sequence tagging models are effective at capturing long bi-directional context, which is crucial automatic punctuation. Automatic Speech Recognition (ASR) production systems, however, constrained by real-time requirements, making it hard to incorporate the right context when decisions. Context within segments produced ASR decoders can be helpful but limiting in overall performance a session. In this paper, we propose streaming approach re-punctuation of output using dynamic decoding windows measure its impact on accuracy across scenarios. The new system tackles over-segmentation issues, improving F0.5-score 13.9%. Streaming achieves an average BLEUscore improvement 0.66 downstream task Machine Translation (MT).