Abstract

People don’t just talk with natural language: sometimes, they talk about it. Dictionaries, language learning resources, scholarly works in linguistics and literature, and even cultural and legal discourse are full of metalanguage explicating linguistic patterns and practices. In this talk I will reflect on metalanguage and what it has to offer for NLP. Are LLMs fluent in metalanguage, and can they provide accurate metalinguistic explanations? I will present case studies looking at two metalinguistically rich genres: (i) online language discussion forums, and (ii) judicial rulings involving language interpretation. Nuanced results from these and other studies indicate the metalinguistic outputs of today’s LLMs cannot always be taken at face value—yet there is enormous potential for systems to better leverage existing metalinguistic data and to assist users with metalinguistic tasks.

Bio

Nathan Schneider is a computational linguist. As Associate Professor of Linguistics and Computer Science at Georgetown University, he leads the NERT lab, looking for synergies between practical language technologies and the scientific study of language, with an emphasis on how words, grammar, and context conspire to convey meaning. He is the recipient of an NSF CAREER award to study NLP vis-à-vis metalinguistic enterprises like language learning, linguistics, and legal interpretation. Recently, he has weighed in on specific interpretive debates in U.S. law; one of these analyses was cited by U.S. Supreme Court justices in a major firearms case. He is active in the NLP community—especially ACL’s SIGANN and SIGLEX—and the Universal Dependencies project; and cofounded the SOLID forum for empirical research on legal interpretation. Prior to Georgetown, he inhabited UC Berkeley, Carnegie Mellon University, and the University of Edinburgh. Apart from annotation scheming and computational modeling, he enjoys classical music and chocolate chip cookies.

Also Available by Zoom:  https://wse.zoom.us/j/96735183473