Page d'accueil

Corrélation géographique des sujets sur Twitter

Across the omnipresent social networks, a cascade of social information is shared at every moment in time, and is increasingly used by the scientific community with novel emerging ways of analysis: data mining. The purpose of this project is thus to travel through some of these data mining features, by enabling analysis of varied subjects using the social platform of Twitter, coupled with a web application: GeoTwit. This application allows users to enter one or two subjects of their choice and provides an expanded geographic filtering of tweets, to visualize a real-time or a (less-expanded) static activity graph/map for these subjects, both on Twitter and on a geographic map, and finally to import and export the results as external files. This tool uses various technologies, among which the Scala language coupled with Play Framework as well as the Twitter's REST and Streaming APIs for the server side, and HTML, CSS and JavaScript for the client side. In preparation for the project, market studies were realized in addition to a thorough analysis of different Twitter's APIs, in order to glean the most optimal approach for the project.

GeoTwit has several potentially interesting use-cases, like a real-time comparison of a population's voting behavior across different parts of a country during national elections, or an analysis of the Twitter's data during a major event, a comparison of certain related subjects (e.g. comparing geographical variation of perception of Java and Scala languages), or lighter hearted topics, for instance, geographic variations in use of the keyword ”vacation”.