Sharing the Collective Intelligence between E-mail Applications
Jean-Baptiste Faddoul, Boris Chidlovskii
This position paper explores the sharing of collective intelligence among different social networks, in particular e-mail applications. The target of such a sharing is to improve the performance of the e-mail categorization, actor role and community detection on the large scale. We present our approach to the cross-application categorization represented as the learning of multiple tasks with different category sets. Our multi-task learning approach is based on the
ensemble principle and extends the conventional Adaboost technique to the multi-task setting.
CIS 2010 - GROUP 2010 Workshop on Collaborative Information Seeking - Sanibel Island, Florida, USA, 7 November 2010