Reproduce “OpenEnded: An Open-Response Speech Corpus for Speaking Proficiency Assessment with Human Annotations and ALM Supervision” 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
Yu-Wen Chen, Eric Zhou, Evelyn Ding, Tianyi Shen, Zhou Yu, Julia Hirschberg
OpenEnded provides a corpus of 9,782 English open-response utterances from Mandarin-speaking learners, rated separately for accuracy, fluency, and prosody. The authors use human ratings for 1,000 test utterances and Gemini-2.0-Flash pseudo-labels for training and development, then train VoxPA. VoxPA reaches a fluency Pearson correlation of 0.699, versus 0.451 for direct model scores, on this test; findings are limited to this collection.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.