From Light to Rich ERE: Annotation of Entities, Relations, and Events

Abstract

We describe the evolution of the Entities, Relations and Events (ERE) annotation task, created to support research and technology development within the DARPA DEFT program. We begin by describing the specification for Light ERE annotation, including the motivation for the task within the context of DEFT. We discuss the transition from Light ERE to a more complex Rich ERE specification, enabling more comprehensive treatment of phenomena of interest to DEFT.

Event Detection and Factuality Assessment with Non-Expert Supervision

“…While event definitions have been proposed in several prior studies, existing approaches vary in how they model various linguistic forms such as nominal events, stative events, generic events, and light verbs (Pustejovsky et al, 2003;Palmer et al, 2005;Meyers et al, 2004;Kim et al, 2009;Song et al, 2015). Even with a formal and precise account of events, training annotators to learn all such linguistic intricacies remains a practical challenge.…”

A Multiform Balanced Dependency Treebank for Romanian

“…Actually, this treebank is also involved in semantics research or in the annotation of entities and events (Song et al, 2015).…”

Proceedings of The 9th Linguistic Annotation Workshop

“…In developing the concept of event hoppers for Rich ERE, we coreference event mentions at the same level of granularity as ACE (i.e., type and subtype match, and sub-events are treated as separate events), but we allow a greater degree of flexibility in the granularity of the arguments that can be participants in coreferenced event mentions than in ACE (Song et al, 2015 wielded a knife)  Argument granularity (18 killed vs. dozens killed) Relaxing the granularity requirements in this way allows annotators to coreference more event mentions that they know refer to the same event. It more closely matches annotator intuitions, and it gives end users a more complete picture of the annotated events and their participants.…”

“…Treebank and ERE guidelines that have been completed for English have been later adapted for other languages as well -for example, Modern Standard Arabic and also dialectal Arabic treebanks (Maamouri and Bies, 2004;Maamouri et al, 2014;Maamouri et al, 2006;, as well as Chinese and Spanish ERE (Song et al, 2015). Clearly, new guidelines are necessary to account for language-specific constructions for each language and annotation task, but developing them based on existing guidelines for another language is a considerable head start.…”