DATE: June 14, 2013
The definition and detection of events have their roots in philosophy and linguistics, with seminal works by Davidson (1969, 1985), Quine (1985) and Parsons (1990), and have long been a subject of study. However, the NLP community has yet to achieve a consensus on the treatment of events, in spite of its critical importance to several areas in natural language processing, such as topic detection and tracking (Allan et al., 1998), information extraction (Humphreys et al., 1997), question answering (Narayanan and Harabagiu, 2004), textual entailment (Haghighi et al., 2005), and contradiction detection (de Marneffe et al., 2008). Most attempts to provide annotation of event coreference have been limited to specific scenarios or domains, as in LDC’s ACE and Machine Reading event annotation, (Humphreys et al., 1997; Bagga and Baldwin, 1999; He, 2007). The recent OnotoNotes annotations include more general event mentions and coreference, but mainly identify coreferences between verbs and nominalizations (Pradhan, 2007). Events are also a crucial element of TimeML, or temporal relation annotation, which have an overlapping but slightly different approach (Pustejovsky, et. al., 2010). Truly comprehensive event detection must encompass the detection of events and their subevents, as well as bridging references (Poesio and Artstein, 2005; 2008). This type of event representation is clearly related to the information available in lexical resources such as PropBank, VerbNet and FrameNet, but goes well beyond anything they currently capture. Bejan and Harabagiu (2010) have recently offered broader event coreference annotation for evaluation purposes, which have been revised and extended by Lee, et.al, (2012). The organizers are themselves involved in event coreference projects for medical informatics and for deep natural language understanding. The time is ripe to bring together interested parties for a serious discussion of appropriate guidelines, resources, and processes for defining and detecting events and their coreferences, and how they should be represented.
In the workshop, James Pustejovsky will be giving the keynote address.
Talk Title: "The Role of Event-based Representations and Reasoning in Language"
Speaker: James Pustejovsky, Brandeis University
The notion of event has long been central for both modeling the semantics of natural language as well as reasoning in goal-driven tasks in artificial intelligence. With the maturation of statistically informed NLP techniques and the increased availability of large linguistically oriented lexical resources, it is time to revisit our theoretical and computational models for events in natural language. In this talk, I outline the challenges for a unified theory of event structure. The demands on such a theory require it to both facilitate the systematic mapping from semantic forms to syntactic representations and support event-based inferences in texts. What emerges is a framework that represents a situation and its participants in terms of subevents, modeled dynamically through time and space. In addition, the theory must identify events as part of larger scenarios and scripts. I discuss recent work in this direction and some models unifiying both representational levels for event-based reasoning.
This is a genuine “working” workshop. The organizers, with the assistance of the program committee, have organized a small shared annotation task on event mention and coreference annotation. The purpose of this annotation is to have all participants look at the principal phenomena of interest and apply their preferred annotation scheme to it. The resulting annotations will be analyzed for agreements and disagreements which will be discussed thoroughly, with examples, in working sessions and panels at the workshop, with the aim of achieving a consensus on the handling of disagreements. Annotation data is available for participants interested in participating, as described below.
The sessions and panels are expected to focus on the following topics:
- Foundations: What are Events? Definition and Recognition
- Coreference: When are Two Events the Same?
- Representation: How Best to Represent Events and Event Groups?
The workshop also has a poster presentation on any of these topics. Please see the Workshop Program Page for more information.
This workshop is endorsed by SIGLEX and SIGSEM.