A Talk by Dr. Miriam Petruck, ICSI Berkeley - FrameNet project
Tue, March 25, 2014 • 3:30 PM - 5:00 PM • BUR 337 Conference Room
The FrameNet project:
Founded in 1997 by Charles J. Fillmore, the FrameNet project (framenet.icsi.berkeley.edu) has been building a unique knowledge base that maps meaning to form through the theory of Frame Semantics (Fillmore 1975, 1985). Started as a computational lexicography project, FrameNet (Ruppenhofer et al. 2010) is the most sophisticated instantiation of Frame Semantics, at the heart of which is the semantic frame, i.e. a schematic representation of an event, situation, state of affairs or object, which facilitates inferences about participants and objects in and across situations and events. Documenting its findings with corpus evidence, the project’s goal is to provide a body of semantically and syntactically annotated sentences from which reliable information can be reported on the combinatorial possibilities of each item analyzed. In addition to presenting and illustrating the fundamental principles of Frame Semantics (Fillmore and Baker 2010), the daily work of FrameNet researchers (Fillmore et al. 2003), and the FrameNet database, the project’s primary product, this talk covers the research that FrameNet has inspired in the natural language processing community, including automatic semantic role labeling (Das et al. 2013), minimally required for a range of more sophisticated NLP tasks, such as information extraction (Mohit and Narayanan 2003), question answering (Narayanan and Harabagiu 2004, Sinha and Narayanan 2005), event extraction (Fillmore et al. 2006) and automatic reasoning (Scheffczyk et al. 2010), as well as automatically induced Frame Semantic lexicons (e.g. Borin et al. 2010). Finally, the talk addresses the importance of FrameNet for linguistic theory and natural language processing application development.
For more information contact Dr. Hans Boas