3 ms·
A massive and continually updated discriminative dataset used to train a language classification model. Build and update continually via mturk or similar and re
by kelseyfrog 4y ago
A massive and continually updated discriminative dataset used to train a language classification model. Build and update continually via mturk or similar and retrain daily. Audit raters (itr) periodically to ensure they're finding signal or at least not disagreeing.