A vast amount of textual web streams is influenced by events or phenomenaemerging in the real world. The social web forms an excellent modern paradigm,where unstructured user generated content is published on a regular basis andin most occasions is freely distributed. The present Ph.D. Thesis deals withthe problem of inferring information - or patterns in general - about eventsemerging in real life based on the contents of this textual stream. We showthat it is possible to extract valuable information about social phenomena,such as an epidemic or even rainfall rates, by automatic analysis of thecontent published in Social Media, and in particular Twitter, using StatisticalMachine Learning methods. An important intermediate task regards the formationand identification of features which characterise a target event; we select anduse those textual features in several linear, non-linear and hybrid inferenceapproaches achieving a significantly good performance in terms of the appliedloss function. By examining further this rich data set, we also propose methodsfor extracting various types of mood signals revealing how affective norms - atleast within the social web's population - evolve during the day and howsignificant events emerging in the real world are influencing them. Lastly, wepresent some preliminary findings showing several spatiotemporalcharacteristics of this textual information as well as the potential of usingit to tackle tasks such as the prediction of voting intentions.
展开▼
