Event Date: Monday, 2 July, 2018
Location: 10-12 and 14-16, Room B3 Palazzo Boilleau
Speaker: Prof. Sebastian Padó (IMS, University of Stuttgart)
Title: Distributional Analysis of Entities
Abstract: Distributional semantics has achieved substantial success in using corpus evidence as the basis for modeling word meaning, disambiguation in context, and semantic relations between words. However, most of the focus of linguistically inspired work has been on common nouns (i.e., categories) while there is little insight into proper nouns (i.e., entities). This tutorial will cover the basics of distributional semantics, modeling (clustering / classification) and then proceed to discussing two recent research questions concerning entities that we investigated: (a) does distributional semantics capture fine-grained encyclopedic information relevant for entities? [Gupta et al. 2015, 2017] (b) how does the semantic relation of instantiation (Lincoln — president) manifest distributionally? [Boleda et al. 2017, Gupta et al. submitted]. The lecture part will be supplemented by a practical session looking at the study data with Jupyter and R.
Sebastian Padó is professor of computational linguistics at Stuttgart University. He received his PhD from Saarland University in 2007 for a thesis on cross-lingual frame semantic modeling. After a postdoctoral stay at Stanford and a junior faculty position in Stuttgart, he become professor of computational linguistics at Heidelberg University in 2010 and returned to Stuttgart in 2013. His core research concerns learning, representing, and processing semantic knowledge (broadly construed) from and in text. Examples include distributional models of linguistic concepts (e.g. compositionality, polysemy), multilingual processing, discourse structure (politeness levels, reported speech, argumentation), and textual entailment. He has taught his kids the scientific concepts of arguing against other’s opinions and justifying their own so well that he doesn’t get to decide where he’s going for holiday any more