Detecting Events and Sentiment on Twitter for Improving Urban Mobility

The streams of tweets from and to the Twitter account of urban transport operators have been considered. A computational module has been designed and developed in order to collect tweets and, on the fly, analyze them to detect some relevant event (e.g. accidents, sudden traffic jams, service interruption, etc.) and/or evaluate possible sentiments and opinions about the quality of service. Events are recognized through a simple word matching while sentiment analysis is performed via supervised learning (Support Vector Machine). The text mining solutions have been developed to work with Italian language; however they could be easily extended to other languages in the case tweets in other languages would be available. This approach has been tested for the urban transportation in Milan (Azienda Trasporti Milano, ATM) in the framework of the TAMTAM project which has developed a technological platform for improving urban mobility by exploiting the large amount of information shared by the users of transportation services through Twitter. Events detected are used by other software modules of the TAM-TAM platform in order to support a more effective travel planning, while sentiment inferred may be used by the transport provider in order to tune the mobility supply to the commuter needs.