Yasemin Altun : Joint learning of Multiple Structured Prediction Tasks
Udo Hahn : What's an Event?
How Ontologies and Linguistic Semantics Shape Upcoming Challenges for Machine Learning Models for Event Extraction
Abstract:
Relation extraction has become the primary concern of researchers involved in information extraction (IE). However, some confusion can be observed what kinds of relations are being considered for IE. They range from binary conceptual relations (e.g., part-whole relations) to complex event structures (e.g., management transitions, interactions of genes, etc.). But what exactly makes an event so special and complex? The answer to this can be found by looking at different forms of event modelling in ontologies and linguistic semantics. In this talk, I will consider various alternatives of event encoding based on ontological and semantic considerations, which differ in their degree of encoding complexity and empirical adequacy. I will then outline upcoming challenges for machine learning models to properly account for event extraction at different levels of representation granularity and quality.
