Skip to main content

Table 4 Equal error rates (EER) for the individual features LR, nr_shorter_1, nr_shorter_5, nr_longer_95, nr_longer_99 and the combinations duration_comb (nr_shorter_1, nr_shorter_5, nr_longer_95, nr_longer_99) and all features, all.

From: Optimizing Automatic Speech Recognition for Low-Proficient Non-Native Speakers

Features EER
LR 14.4%
nr_shorter_1 27.3%
nr_shorter_5 27.4%
nr_longer_95 35.8%
nr_longer_99 38.5%
duration_comb 25.3%
all 10.3%