2020-21 Distinguished Computational Linguistics Lecture

2020-2021 Distinguished Computational Linguistics Lecture

Register here: http://bit.ly/3c4sXY3

Presented by: Language Science Program at RIT
https://www.rit.edu/cla/languagescience/

Title: Describing word meaning is like nailing jelly to the wall. Can embeddings help?

Speaker: Katrin Erk, Professor of Linguistics, The University of Texas at Austin

Abstract: Word meaning in context is notoriously flexible. Maybe there isn't even such a thing as "word senses." If that is so, then word embeddings may actually be an excellent way to characterize word meaning for lexical semantics: They can be automatically computed from corpus data, and can then be used to analyze the word meanings that appear in the corpus, synchronically or even diachronically. They can characterize differences in token meaning without resorting to word senses -- if we can figure out how to use them in the right way. In this talk, I will describe work that we have done on this: experiments on how to use embeddings for lexical semantics (including some lexical BERTology), annotation tasks to better understand what word meaning is, and some attempts to integrate word embeddings in formal semantics.

To request an interpreter, visit myaccess.rit.edu


Contact
Zhong Chen
Event Snapshot
When and Where
March 31, 2021
3:00 pm - 4:00 pm
Room/Location: http://bit.ly/3c4sXY3
Who

This is an RIT Only Event

Interpreter Requested?

No