From event tracking to event modelling : understanding as a paradigm shift

Abstract

In 1998, Topic Detection and Tracking (TDT), or event tracking, was only two years old. By then, however, event tracking’s pioneers had already realized that for algorithms to accomplish their task, to detect and track events in the news media accurately, they had to understand events. Yet it has taken event tracking the decades since then to resolve what it means to understand events and how. Simple interpretations of understanding often failed; the rest never fulfilled the potential of event tracking. Without understanding, event tracking continues to struggle with the same, early problems, which social media aggravated and contemporary applications of events compounded. Therefore in this position paper, we argue that now, more than ever, event tracking needs event understanding. By comparing event tracking’s interpretations of what it means to understand, we demonstrate how early understanding only failed event tracking because it remained too simple in its reliance on linguistics. We adopt a different definition of events, a structured and semantic description that revolves around Who did What, Where and When. Furthermore, we propose how event tracking can fill in the event structure automatically and ahead of time. In the end, we show that understanding remains a complex problem, but event tracking can find in event knowledge a new purpose and its much-awaited paradigm shift.