Web and social medias are an incredible platform for the exchange of information, adopted by a large number of users.
The subject of this thesis concerns the domain of opinion mining on those platforms, to automatically detect opinion trends on given subjects. More than the simple attribution of a polarity on a piece of text, a finer analysis is often required to have better caracterization of opinion (finding which object or which precise feature of the object is covered by the opinion).
Another characteristic of social medias is the structured organization of users' relationship among communities (« friends » on Facebook, « followers » and « re-tweets » on Twitter, comments on different platforms). This structuration may be used to find information about users' social importance, like their popularity or reputation.
This thesis aims to combine these elements to produce a consolidated synthesis of sentiment trends about subjects or people according to their spatial and temporal context.
La Nuit de l’Info is a national competition gathering students, teachers and industries to develop a web application during one night.