When: 2014-08-27 at 09:00-10:00
Where: E:2116, E-house, Ole Römers väg 3, Lund
Speaker: Pierre Nugues /Marcus Klang
Abstract: Question answering systems have been the focus of an intensive media coverage when IBM demonstrated a system that could answer questions better than any human being in the US Jeopardy quiz show. Such systems are based on the application of semantic processing techniques to large collections of text such as Wikipedia or the New York Times archive and an automatic knowledge extraction. This new generation of question answering systems is made possible as gigantic amounts of text are becoming machine readable and will probably play a significant role in the future of information technology.
In this seminar, we will review the architecture of question answering systems and describe their main components. We will also detail Hajen, a question-answering system we are developing for Swedish at LTH and show demonstrations of our current prototype. In our presentation, we will examine more specifically one component: the named entity disambiguation. This component enables the system to associate a word or a phrase referring to a person, country, product, organization etc., to a unique identifier given the context of the word.
Early version of Hajen: http://semantica.cs.lth.se:8888/sv/query/qa