You are here
NewsReader Workshop: Perspectives on Events
In the framework of the NewsReader European project, we have organized a one-day workshop with invited speakers trying to address issues and questions around the topic of "Events".
Since one of the goals of NewsReader is to extract information on events from large streams of news and build structured indexes for decision making in the financial and economical domain, it is crucial to define what an event is. Although 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, the research community has yet to achieve a consensus on what is an event and how it should be modeled. The goal of this workshop is to bring together interested parties from different research fields for a discussion of appropriate guidelines, resources, and processes for defining and detecting events.
Note: The workshop is open to students and researchers, attendance is free, no registration needed.
Sara Tonelli and Rachele Sprugnoli
- 14.30 - 14.45 Welcome
- 14.45 - 15.30 Achille Varzi (Columbia University) "Events: What They Are and What They Are Not"
- 15.30 - 16.15 Luciano Serafini (DKM-FBK) "Events in Logic"
- 16.15 - 16.45 Coffee break
- 16.45 - 17.30 Nicola Guarino (LOA) "Episode-centric conceptual modelling"
- 17.30 - 18.15 Bernardo Magnini (HLT-FBK) and Tommaso Caselli (TrentoRise - IBM) "Events in Computational Linguistics"
- 18.15 - 18.30 Closing
Achille Varzi, "Events: What They are and What They are Not"
Are events entities of a kind? If so, what are they? How do they differ from facts, states of affairs, tropes? How are they related to objects and properties? What are their identity and individuation criteria? Are there also negative events, such as failures, omissions, avoidances? Alas, none of these questions admits of a unique, uncontroversial answer, so I won't try to give any. But we can draw a map of the options. And looking at the map we can at least assess the degree to which some options are, all things considered, or relative to certain purposes, better than others.
Luciano Serafini, "Events in Logic"
Formal representation of knowledge about events is one of the most important research areas of AI and CS. Being able to automatically reason about events that actually happen or may happen is extremely relevant for many areas of AI and CS, such as planning, natural language processing, formal verification of programs, decision making, ontologies and semantic web. For this reason, many logics have been developed in the last 40 years with the objective of formally representing events. In this talk, I will try to give a (necessarily partial) overview of the different approaches and their main characteristics and applications.
Nicola Guarino, "Episode-centric conceptual modelling"
Most conceptual modeling and knowledge representation schemes focus on relations. The Entity-Relationship approach is a paradigmatic example in this respect. However, in many cases, when we say that a relationship *holds* (say, John is married with Mary) there is something which *occurs* (or "perdures") in a certain interval of time. Technically speaking, this entity is a *perdurant*, in DOLCE's terms. We may call it an event for short, in the most general understanding of this term, but I tend to prefer the term *episode*, for the reasons I will illustrate in my talk (according to OALD, an episode is "an event, a situation, or a period of time that is important or interesting in some way").
My main point will be that, whenever there is an episode which corresponds to a particular relationship, it is very useful to model such episode explicitely, putting it in the domain of discourse. I will investigate the various cases when a relationship corresponds to an episode, and illustrate a number of open ontological issues concerning episodes and their participants.
Bernardo Magnini and Tommaso Caselli, "Events in Computational Linguistics"
In this talk we briefly review some relevant linguistic features of events and show how such features have been addressed in Computational Linguistics, pointing out recent approaches, resources and initiatives in the community. Specifically, we will focus on the following four phenomena: (i) recognizing different kinds of eventualities; (ii) automatically recognize the structure of an event, highlighting its subcategorization frame, its thematic roles and their selectional preferences; (iii) the interpretation of an event through its locutionary properties, aiming at the automatic identification of factivity; (iv) finally, we report on approaches addressing the identification of the temporal relations among events expressed by verbs.