Found in Translation
We present a complete working system that gathers multilingual news items from the Web, translates them into English, categorises them by topic and geographic location and presents them to the final user in a uniform way. Currently, the system crawls 560 news outlets, in 22 different languages, from the 27 European Union countries. Data gathering is based on RSS crawlers, machine translation on Moses and the text categorisation on SVMs. The system also presents on a European map statistical information about the amount of attention devoted to the various topics in each of the 27 EU countries. The integration of Support Vector Machines, Statistical Machine Translation, Web Technologies and Computer Graphics delivers a complete system where modern Statistical Machine Learning is used at multiple levels and is a crucial enabling part of the resulting functionality.