Last week Lora Aroyo gave a keynote “Dial E for Events” at the i-semantics2012 conference. The focus of the talk were events and an approach, based on the theory that the disagreement among humans about events constitute a natural state, where we can harness this disagreement through crowdsourcing to bring a new kind of meaning to events.
Objects, like people, locations, and various other types of named entities, are often easy to detect in language and present on the semantic web. Without events, however, they lack meaning. Assigning roles to objects in events is a step towards bringing them meaning, but the detection and representation of events is much harder than for objects; they are typically not named and humans have difficulty identifying them and distinguishing their boundaries, as well as linking and ordering them consistently. Many event-centric approaches in NLP have attempted to “fix” the problem of human disagreement regarding events by over-specifying their semantics for isolated tasks, but this leads to brittleness and lack of coverage.
Here are the slides: