CompSci Colloquium: Jinho Choi
ABSTRACT: In this talk, Dr. Choi introduces the “Character Mining” project that aims to develop innovative models for machine comprehension on multiparty dialogue, and presents his latest work on two tasks, character identification and question answering. Character identification is an entity linking task that identifies each personal mention as a certain character in a dialogue. A novel transition-based coreference resolution algorithm is designed to handle both singular and plural mentions, and the agglomerative convolutional neural network is used to develop entity liking models for this task. For question answering (QA), a new dataset is created, FriendsQA, that tackles span-based QA where the evidence document is a dialogue. The state-of-the-art contextualized embedding approach, BERT, is used to train advanced QA models for this task. Additionally, Dr. Choi gives an overview of his ongoing project on conversational AI called “Emora” that aims to bring the true meaning of social chatbot to the community.
Jinho Choi is an assistant professor of the Department of Computer Science, the Institute of Quantitative Theory and Methods, and the Program of Linguistics at Emory University. He obtained B.A. in Computer Science and Mathematics (dual degree) from Coe College in 2002, M.S.E. in Computer and Information Science from the University of Pennsylvania in 2003 (advisor, Mitch Marcus), Ph.D. in Computer Science and Cognitive Science (joint degree) from the University of Colorado Boulder in 2012 (advsior, Martha Palmer), and post-doctoral training at the University of Massachusetts Amherst in 2014 (advisor, Andrew McCallum). He was a full-time lecturer in the Department of Computer Science at the Korea Military Academy from 2004 to 2007 while he served his military duty in South Korea. Dr. Choi is the founder and the director of the Natural Language Processing Research lab at Emory University.
Dr. Choi has been active in the field of Natural Language Processing (NLP). He has presented many state-of-the-art NLP models that automatically derive various linguistic patterns and structures from free text. These models are publicly available through the cloud-based NLP platform called ELIT, the successor of NLP4J and ClearNLP, that Dr. Choi has created to promote academic and industrial research. Since he came to Emory, Dr. Choi has introduced novel machine comprehension tasks to identify personal entities and infer explicit and implicit contexts in multiparty dialogue, which can be used to build question answering systems on human conversion. For the application of his research, Dr. Choi has developed innovative models to classify severity levels on radiology reports using deep neural networks and detect early stages of Alzheimer’s disease using meta-semantic analysis, which show similar accuracy as human experts in those domains.
Thursday, October 17 at 3:30pm to 4:30pm
Engineering Center, ECCR 265
1111 Engineering Drive, Boulder, CO 80309