Reproduce “Speech event detector: Timestamped detection of non-verbal vocalizations and background sounds for rich speech transcription” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ayush Kumar Bar, Promila Ghosh
Speech Event Detector adds timestamped sound tags alongside recognized words. Its 5.8-million-parameter convolutional–Transformer model predicts 54 sound classes every 40 ms, using labels combined from seven sources. On 9,367 held-out events, its sensitive profile reaches 0.491 macro recall and 0.302 macro precision. Thresholds were tuned on the evaluation clips, so these scores are optimistic and do not establish performance elsewhere.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.