Tracking Out-of-date newspaper Articles
Frédérik Cailliau, Aude Giraudel, Béatrice Arnulphy · 2009
Abstract. Local newspapers rely on their local correspondents to bring you the freshest news of your home town. Some of the articles written by these correspondents are not published immediately, but are put aside to be placed in a later edition. One of the challenges the editor in chief is confronted with, is to publish only up-to-date information about a local event. In this paper we present a system that tracks out-of-date newspaper articles to prevent their publishing. It firstly dates the event the article is talking about. The date detection grammar is written in a single but complex finite state automaton based on linguistic pattern matching. Secondly, it computes an absolute date for each relative date. A freshness score can then be deduced from the time difference between the extracted and the publication date. The system has been tested on several French local newspaper corpora. Baseline for the temporal extraction is our standard date extraction that was developed for general purposes.