Reproduce “Subphonetic Acoustic Modeling via Optimal Transport for Pronunciation Assessment” 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
Haopeng Geng, Jiun-Ting Li, Daisuke Saito, Nobuaki Minematsu
The authors propose a pronunciation-assessment model that represents each phone with ordered subphonetic states and trains dense frame-to-state evidence using optimal temporal transport classification. In an L2-ARCTIC ablation at 20-ms resolution, this criterion raises segmentation R-value from 47.60 to 68.34 and lowers timestamp error from 78.14 to 40.64 ms; the result is limited to that setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.