B08: Hierarchical frame induction via probabilistic models

Funding period: 01.07.2015 - 30.06.2019

Project B08 aims at the unsupervised induction of frames for events described by verbs, using techniques from probabilistic frame modeling and from machine learning. The outcome should not only identify predicate-argument relations but also provide a rich characterization of these relations via hierarchically ordered semantic roles. This allows then the induction of an event type hierarchy that will be checked against and provide evidence for the event types assumed in the linguistic CRC projects and in the literature. In a second step, B08 also seeks to learn relations between the frames of verbs that are subject to alternations. This goal will be pursued in a semi-supervised approach.