Image-event relationship analysis for finding and attaching representative images to event nodes of a temporal event map

Abstract

As the information burst fast in 21th century, peoples are easily out of track of an event progress (e.c. news event progress). Currently, Temporal Event Map (TEM) is introduced by Li et al. [3] to give a whole view on an event progress. However, the TEM is presented in text form with edges between nodes indicating event relationship. It is not readable friendly for people to understand what the event nodes talk about. Therefore, in order to improve its readability and saving reader’s time, the problem of finding and attaching representative image of event node is formulized in this study and some approaches are proposed to deal with the problem. Image Event Relationship System model is proposed and we defined three kinds of correlations which are text-image correlation (ITC), event-image correlation and event node-event node correlation. We also proposed methods to handle these correlations and thus representative image for event nodes of a TEM can be found.