A Common Measure of Communication for Speech Brain-Computer Interfaces
The paper proposes OVMI as a common information scale for comparing speech BCI systems built under different constraints.
The authors argue that accuracy and WER can make systems look stronger than they are when those scores only cover supported words. OVMI measures how much intended speech a decoder can convey against a reference distribution of words a user may want to say. Using it, they compare existing systems, show vocabulary-versus-accuracy trade-offs, and report up to 16.3% relative accuracy improvement from OVMI-based vocabulary selection across three speech domains. Source: ArXiv · AI/CL/LG's note
The authors argue that accuracy and WER can make systems look stronger than they are when those scores only cover supported words. OVMI measures how much intended speech a decoder can convey against a reference distribution of words a user may want to say. Using it, they compare existing systems, show vocabulary-versus-accuracy trade-offs, and report up to 16.3% relative accuracy improvement from OVMI-based vocabulary selection across three speech domains. Source: ArXiv · AI/CL/LG's note
score 4