Marie-Francine Moens
http://www.cs.kuleuven.be/~sien/
Associate Professor
Department of Computer Science
Universiteit Leuven
Belgium
pour son tutorial : "Mining Opinions in Blogs, The Good, the Bad and the Ugly"
Blogs offer a variety of opinions on politics, consumer products, persons and other subjects. We cannot neglect this wealth of information. No wonder that there are multiple research and commercial efforts to make this information accessible. In a first part the lecture gives an overview of state of art technologies, their results and their applications. In a second part we describe and discuss our own research in opinion mining and focus on the following problems. As blogs interweave many different opinions expressed towards a variety of targets, it is important to correctly attribute a certain opinion to the right subject of discussion. Moreover, the language used in blogs is very diverse demanding automated methods for efficiently selecting examples when acquiring or annotating the necessary opinion patterns (e.g., use of active learning techniques). Related to the foregoing is that blog languages diverge largely from standard language and evolve continuously. We need here very adaptive mining systems. In a third part of the lecture, we go deeper into the current phenomenon of spamming and manipulation of opinions and how text mining could be of help in order to protect sensitive users such as children. Throughout the lecture, we point to many promising research avenues.