Reproduce “SignAligner: Dual-Granularity Span Generation for Sign Language Alignment” 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
Jongbhin Woo, Youngjoon Jang, Joon Son Chung
SignAligner trains one model to predict start and end times for subtitle sentences or queried English words, using subtitle spans and mined word spans, then refines predictions with overlap rewards on manually aligned examples. On BOBSL’s manually aligned test set, it reaches 66.68 mean F1 at 0.50, among compared systems without external sign-video resources, for broadcast BSL with audio-aligned priors.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.