Reproduce “From 2.1M Hours of Raw Broadcast Audio to Finnish Speech SSL: Confidence Filtering and Architectural Analysis at Scale” 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
Yaroslav Getman, Tamás Grósz, Tommi Lehtonen, Mikko Kurimo
The pipeline turns a 2.1 million hours Finnish broadcast archive into a 471,000-hour pretraining set using speech detection, transcript deduplication, and ASR confidence as a proxy for audio quality. In controlled tests, confidence filtering at roughly 72% retention lowered WER on three of four validation sets versus language-identification filtering. The result applies to the tested Finnish data, models, and benchmarks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.