Multi-hop QA, Modeling Events, and Translation in Neural Models

Explore cutting-edge research in natural language processing through this 54-minute conference talk by Nirajan Balasubramanian from Stony Brook University. Delve into three key areas: repurposing pre-trained neural entailment models for multi-hop question answering, creating datasets that ensure multi-hop reasoning, and decomposing large QA models for mobile devices. Examine methods for learning structured latent spaces to better control event sequence modeling and generation. Gain insights into modeling target side syntax for machine translation. Learn about the speaker's work at the LUNR lab, focusing on various NLP problems including QA, event knowledge, language generation, and efficient NLP for mobile devices.