1 paper · 1 filter
Gaurab Baral, Mimansak Nepal, Grishma Basnet
Many sentiment datasets use one annotator for each item, so the test labels show the conventions of that person. We build a Nepali benchmark from two annotators who labelled the sa…