Reproduce “SPARC: Spoken Profile Authorship Recognition Corpus” 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
Efthymios Tsaprazlis, Bowen Yi, Tiantian Feng, Junghwan Kim, David Jurgens, Shrikanth Narayanan
SPARC tests whether transcript excerpts retain speaker-linked information after acoustic cues are removed, using pairwise verification rather than identification among a population. It compares reference and probe texts across episodes as topic controls. On NaturalVoices cross-episode trials, GPT-5-mini reaches an EER of 0.285 versus 0.351 for SentenceBERT; episodes are imperfect topic proxies, and the cues driving linkage remain unknown.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.