Real-time event extraction for driving information from social sensors

Abstract

Car navigation systems, which are a part of technologies of Intelligent Transportation System (ITS), support drivers in driving efficiently by showing the shortest routes and some driving information such as that related to traffic jams and weather reports, that are collected from physical sensors and news media. This paper presents a proposal of a method to extract real-time driving information using social media as a type of 'social sensor'. This is a brand-new approach to acquiring valuable information for drivers from social media. Our method is useful to provide driving information in countries where current ITS deployment is poor. It also provides drivers detailed information related to the current traffic situation. Our research comprises three parts. First, we extract driving information from social media using text-based classification methods. Second, because geographical coordinates are necessary to note where the driving information had occurred, we incorporate a method to transform geographically related terms into geographical coordinates. Finally, we develop a system to provide information about important events for drivers and to evaluate results of event extraction through comparison with information available from current media sources commonly found in most cars. Our proposed system can collect tweets referring to heavy traffic from Twitter with about 0.87 precision and can extract location information from those tweets with 0.85 precision.