A methodology for traffic-related Twitter messages interpretation

This paper addresses the problem of interpreting tweets that describe traffic-related events and that are distributed by government agencies in charge of road networks or by news agencies. Processing such tweets is of interest for two reasons. First, albeit phrased in natural language, such tweets use a much more regular and well-behaved prose than generic user-generated tweets. This characteristic facilitates automating their interpretation and achieving high precision and recall. Second, government agencies and news agencies use Twitter channels to distribute real-time traffic conditions and to alert drivers about planned changes on the road network and about future events that may affect traffic conditions. Hence, such tweets provide exactly the kind of information that proactive truck fleet monitoring and similar applications require. The main contribution of the paper is an automatic tweet interpretation tool, based on Machine Learning techniques, that achieves good performance for traffic-related tweets distributed by traffic authorities and news agencies. The paper also covers in detail experiments with real traffic-related tweets to access the precision and recall of the tool.