Abstract-Concrete Relationship Analysis of News Events Based on a 5W Representation Model

In a follow-up news article, description of previous news events may be abbreviated or summarized. This feature makes news article difficult to understand if the reader has no knowledge about the previous events. In such a case, providing concrete and detailed descriptions is helpful. In this paper, we propose a five element, who, what, whom, when, and where (5W) model and extraction method with completion functionality. With this model, a news event is represented using these 5Ws. To discover abstract and concrete descriptions of a given event, we propose the novel concept of abstractiveness based on this model. The abstractiveness of an event description is defined based on the difficulty of imagining and identifying that event. Currently, we estimate the abstractiveness of an event by considering the abstract levels and comprehensivity of its 5Ws to identify that event. We also propose a method for estimating the abstractiveness of an event and analyzing the abstract-concrete relationships between news events based on the 5W model. The experimental results indicate that our model, concept, and method are effective for extracting a concrete event description.