Efficient Punctuation Restoration via Weighted Lookahead Scoring for Streaming ASR

Arxiv CS.CL (NLP/language models)

The paper proposes a weighted lookahead scoring method for punctuation restoration in streaming ASR systems, enabling online decisions with limited future context. The approach balances accuracy and latency by dynamically weighting lookahead information.

Arxiv CS.CL (NLP/language models)·Sungmook WooHyungu KangChanwoo Kim

6 days ago