Please join us for a seminar on 3rd December (Tuesday) from 1-2 pm by Marzena Karpinska, a PhD student at the University of Tokyo Linguistics department.
Model evaluation: What can we do better?
Tuesday, 3rd December 2019, 1:00pm to 2:00pm.
ABSTRACT- “Progress in NLP is driven by tasks: we say that a new model is better than previous models if it beats them in a certain set of benchmarks. Clearly, this approach depends highly on the quality of the evaluation sets, and these vary not only by task but also by language. This talk will explore the problems that stem from modeling resources for languages other than English, and from machine translation often used to create such datasets.”
BIO- Marzena Karpinska is currently finishing her Ph.D. in Linguistics at the University of Tokyo, Department of Language and Information Sciences. She received her Master’s degree in Japanese Linguistics from the University of Warsaw and in Linguistics from the University of Tokyo. Her research interests lie at the intersection of natural language processing, theoretical linguistics, and language acquisition.
Faculty Host:Mohit Iyyer