Multi-task learning across three Duolingo language datasets improves word-level answer prediction in low-resource settings, roughly matching 10x larger single-task training sets.
In: 2013 IEEE Inter- national Conference on Acoustics, Speech and Signal Processing
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Multi-task Learning for Low-resource Second Language Acquisition Modeling
Multi-task learning across three Duolingo language datasets improves word-level answer prediction in low-resource settings, roughly matching 10x larger single-task training sets.