Wednesday, 27th March - 11am Valentina Pyatkin : Seminar

Title: Underspecified Language in Context

 

Abstract: 

In this talk I will discuss the role of context and how it is able to restrict and specify implicit, ambiguous and underspecified language. I will showcase the role of context using two different domains. 

First, I will talk about superlatives and how their semantic interpretations are often underspecified. Superlatives are used to single out elements with a maximal/minimal property. Semantically, superlatives perform a set comparison: something has the min/max property out of a set. As such, superlatives provide an ideal phenomenon for studying implicit phenomena and discourse restrictions. While this comparison set is often not explicitly defined, its (implicit) restrictions can be inferred from the discourse context the expression appears in. 

Next, I will talk about social and moral situations and how additional context is able to flip the moral judgment associated with such situations. ClarifyDelphi is an interactive system that learns to ask clarification questions (e.g., why did you lie to your friend?) in order to elicit additional salient contexts of a social or moral situation. We posit that questions whose potential answers lead to diverging moral judgments are the most informative. Thus, we propose a reinforcement learning framework with a defeasibility reward that aims to maximize the divergence between moral judgments of hypothetical answers to a question.

 

Bio:

Valentina Pyatkin is a Postdoctoral Researcher at the Allen Institute for AI and the University of Washington, working with Prof. Yejin Choi and the Mosaic team. She completed her PhD in Computer Science at the NLP lab of Bar Ilan University, supervised by Prof. Ido Dagan and Prof. Reut Tsarfaty. She was also a visiting PhD student at UW NLP and had the pleasure of interning twice at AI2, where she was awarded the AI2 Outstanding Intern of the Year award. Previously she also completed a research internship at Google and obtained an MSc from the University of Edinburgh and a BA from the University of Zurich

Mar 27 2024 -

Wednesday, 27th March - 11am Valentina Pyatkin : Seminar

This event is co-organised by ILCC and by the UKRI Centre for Doctoral Training in Natural Language Processing, https://nlp-cdt.ac.uk.

IF G.03