Investigating representations of verb bias in neural language models

linguists need to be careful about assuming verbs fall into “alternating” and “non-alternating” classes: pic.twitter.com/jaThjFBesz

— Adele Goldberg (@adelegoldberg1) September 16, 2020

Models do reasonably well at capturing length and definiteness effects (‘cept what’s with BERT preferring lexical over pronoun recipients?): pic.twitter.com/0virtvykQt

— Adele Goldberg (@adelegoldberg1) September 16, 2020