PhD defense of Dennis Diefenbach

PhD defense of Dennis Diefenbach

at 2:00 PM

Room F021
Building F
Laboratoire Hubert Curien
18 rue Professeur Benoît Lauras
42000 Saint-Etienne

"Question Answering over Knowledge Bases"

Abstract

Question Answering (QA) is a field in computer science, which is concerned about building a system, which can automatically answer a given question posed by user in natural language. This thesis concentrate on QA over knowledge bases (KB), structured collection of information.

This thesis provides several contributions. The main contribution consists of an algorithm to construct QA systems over KBs. This algorithm improves previous QA approaches in terms of following characteristics: it can be easily adapted to new languages as well as to new KBs, it supports both keyword questions and full natural language questions, and it is robust to malformed questions. The performance of our system is compared over popular benchmarks. We beat most of the existing systems in terms of response time and F-measure, while we are not only limited to English and to a single KB.

The other contributions of the thesis relate to the interactions with the users, to the selection of additional information to be presented with the answer, and to the architecture for a modular QA system construction. All the contributions of this thesis are integrated in an online demo that is available under http://www.wdaqua.eu

Committee

Kamal Singh, Université Jean Monnet, Co-supervisor
Gerhard Weikum, Max Plank Institute in Saarbrücken, Reviewer
Elena Cabrio, University of Nice Sophia Antipolis, Reviewer
Andreas Both, DATEV Corp., Examiner
Brigitte Grau, Ecole nationale supérieure d’Informatique pour l’industrie et l’entreprise, Examiner
Vanessa Lopez, IBM Ireland Research

The defense will be done in English