Following the success of the two workshops on Computing News Storylines (CNewsStory, ACL 2015, CNewsStory, EMNLP 2016), the previous four editions of the EVENTS workshops (NAACL 2013, 2014, 2015, 2016), and the strong connections between these two initiatives, the organisers of the two series have decided to join efforts and propose this new workshop: Events and Stories in the News.
Traditionally, event structures have always been at the heart of linguistics. Nevertheless, NLP solutions seem to struggle with notions such as event detection, event identity and coreference, and causal, topical, temporal, and spatial relations between events.
Humans, on the other hand, easily notice salient events, such as changes in the world, make reference to these events and their participants through linguistic expressions, and construct storylines using only the pertinent aspects. Such narrative structures are at the heart of information sharing, as is exemplified by the structure of news articles. But it remains extraordinarily difficult to detect them automatically, let alone decide on pertinence for automated story construction from rich event representations.
Today’s massive news streams demand multidimensional, multimodal, distributed representations of events and narrative structures that take the connections between all relevant elements involved in a “story”. Currently, most work on cross-document temporal processing focuses on linear timelines, assuming a simple chronological ordering of events in time. However, not every timeline necessarily forms a good and useful storyline.