Abstract

As NLP moves into tasks requiring deeper language understanding, inferencing, and reasoning, knowledge-based resources are being increasingly called upon to support and supplement probabilistic and other data-driven methods. PropBank, VerbNet, FrameNet, the Rich Entities Relations and Events, and Abstract Meaning Representation projects are all valuable NLP resources for English that share the goal of associating semantic information with propositions of natural language.  However, each differs in the level and nature of semantic detail represented, which is especially apparent when comparing the word sense distinctions made by each project.  In this talk, I will explore the strengths and weaknesses of these resources, including the somewhat limiting compositional, lexicosemantic approach that has been taken in developing many of them.  I describe recent efforts to expand annotations to capture the meaning of semi-compositional constructions.  I will also introduce the Rich Event Ontology, which provides an independent conceptual infrastructure unifying these resources, thereby allowing users to take advantage of each resource’s unique contributions while overcoming some of their individual gaps and weaknesses. Notably, the ontology also contributes another level of relational information, in the form of event-to-event causal and temporal relations, which is not found in any of the individual repositories.

Biography

Claire Bonial is a U.S. Army Research Laboratory computational linguist, specializing in the murky world of event semantics. In her efforts to make this world computationally tractable, she has collaborated on a variety of NLP infrastructure projects, such as PropBank, VerbNet, and Abstract Meaning Representation, each of which provides a unique representation of event semantics.  A focal contribution to these annotation projects has been her theoretical and psycholinguistic research on both the syntax and semantics of English Light Verb Constructions (e.g., take a walk, make a mistake).  As a student and researcher in Martha Palmer’s lab at the University of Colorado Boulder (Ph.D. in Linguistics and Cognitive Science, 2014), Bonial served in a multitude of roles on these projects, including annotation management (from graduate students to Mechanical Turkers) and guideline development/expansion grounded in theoretical linguistic research.  Bonial began her current position in the Computational and Information Sciences Directorate of the Army Research Laboratory in 2015.  Since joining ARL, she has expanded research interests to multi-modal, text/video representations of events as well as human-robot dialogue.