Visually Grounded Language Learning

Typically, visually-grounded language learning systems only accept feature data about objects in the environment that are explicitly mentioned, whether through annotation labels or direct ref- erence through natural language. We …...