{"642100":{"#nid":"642100","#data":{"type":"event","title":"ML@GT Virtual Seminar: Ellie Pavlick, Brown University","body":[{"value":"\u003Cp\u003EML@GT is hosting a virtual seminar featuring Ellie Pavlick from Brown University.\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Ca href=\u0022https:\/\/primetime.bluejeans.com\/a2m\/register\/esbdzzaf\u0022\u003E\u003Cstrong\u003ERegistration is required.\u003C\/strong\u003E\u003C\/a\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\r\n\r\n\u003Ch3\u003EYou\u0026nbsp;can\u0026nbsp;lead\u0026nbsp;a\u0026nbsp;horse\u0026nbsp;to water...: Representing vs. Using Features in Neural\u0026nbsp;NLP\u003Cbr \/\u003E\r\n\u0026nbsp;\u003C\/h3\u003E\r\n\r\n\u003Ch4\u003EAbstract\u003C\/h4\u003E\r\n\r\n\u003Cp\u003EA\u0026nbsp;wave\u0026nbsp;of\u0026nbsp;recent work has sought to understand how pretrained\u0026nbsp;language\u0026nbsp;models work. Such analyses have resulted in two seemingly contradictory sets\u0026nbsp;of\u0026nbsp;results. On one hand, work based on \u0026quot;probing classifiers\u0026quot; generally suggests that SOTA\u0026nbsp;language\u0026nbsp;models contain rich information about\u0026nbsp;linguistic\u0026nbsp;structure (e.g., parts\u0026nbsp;of\u0026nbsp;speech, syntax, semantic roles). On the other hand, work which measures performance on\u0026nbsp;linguistic\u0026nbsp;\u0026quot;challenge sets\u0026quot; shows that models consistently fail to use this information when making predictions. In this talk, I will present\u0026nbsp;a\u0026nbsp;series\u0026nbsp;of\u0026nbsp;results that attempt to bridge this gap. Our recent experiments suggest that the disconnect is not due to catastrophic forgetting nor is it (entirely) explained by insufficient training data. Rather, it is best explained in terms\u0026nbsp;of\u0026nbsp;how \u0026quot;accessible\u0026quot; features are to the model following pretraining, where \u0026quot;accessibility\u0026quot;\u0026nbsp;can\u0026nbsp;be quantified using an information-theoretic interpretation\u0026nbsp;of\u0026nbsp;probing classifiers.\u003Cbr \/\u003E\r\n\u0026nbsp;\u003C\/p\u003E\r\n\r\n\u003Ch4\u003EAbout Ellie\u003C\/h4\u003E\r\n\r\n\u003Cp\u003EEllie Pavlick is an Assistant Professor\u0026nbsp;of\u0026nbsp;Computer Science at Brown University where she leads the\u0026nbsp;Language\u0026nbsp;Understanding and Representation (LUNAR) Lab. She received her PhD from the one-and-only University\u0026nbsp;of\u0026nbsp;Pennsylvania. Her current work focuses on building more cognitively-plausible models\u0026nbsp;of\u0026nbsp;natural\u0026nbsp;language\u0026nbsp;semantics, focusing on grounded\u0026nbsp;language\u0026nbsp;learning and on sample efficiency and generalization\u0026nbsp;of\u0026nbsp;neural\u0026nbsp;language\u0026nbsp;models.\u003C\/p\u003E\r\n","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":"","field_summary_sentence":[{"value":"ML@GT is hosting a virtual seminar featuring Ellie Pavlick from Brown University. "}],"uid":"34773","created_gmt":"2020-12-14 15:14:05","changed_gmt":"2021-03-09 16:14:59","author":"ablinder6","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2021-03-24T13:15:00-04:00","event_time_end":"2021-03-24T14:15:00-04:00","event_time_end_last":"2021-03-24T14:15:00-04:00","gmt_time_start":"2021-03-24 17:15:00","gmt_time_end":"2021-03-24 18:15:00","gmt_time_end_last":"2021-03-24 18:15:00","rrule":null,"timezone":"America\/New_York"},"extras":[],"groups":[{"id":"47223","name":"College of Computing"},{"id":"37041","name":"Computational Science and Engineering"},{"id":"1299","name":"GVU Center"},{"id":"589608","name":"Machine Learning"},{"id":"576481","name":"ML@GT"},{"id":"431631","name":"OMS"},{"id":"50877","name":"School of Computational Science and Engineering"},{"id":"50875","name":"School of Computer Science"},{"id":"50876","name":"School of Interactive Computing"}],"categories":[],"keywords":[],"core_research_areas":[],"news_room_topics":[],"event_categories":[],"invited_audience":[{"id":"78761","name":"Faculty\/Staff"},{"id":"177814","name":"Postdoc"},{"id":"78771","name":"Public"},{"id":"174045","name":"Graduate students"},{"id":"78751","name":"Undergraduate students"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp\u003EAllie McFadden\u003C\/p\u003E\r\n\r\n\u003Cp\u003Eallie.mcfadden@cc.gatech.edu\u003C\/p\u003E\r\n","format":"limited_html"}],"email":[],"slides":[],"orientation":[],"userdata":""}}}