I. Introduction
Automatic Question Generation (QG) has a wide body of applications, for example to generate quizzes for entertainment purpose such as the famous Trivial Pursuit board game, or online trivia applications. QG also has applications in the educational field, in order to automatically assess the understanding of students [1], [2]. The rise of large Knowledge Bases (KB) such as DBpedia, Yago, Freebase and Wikidata has renewed the interest of the community for Question Answering (QA) systems [3], [4]. The lack of standard datasets to assess these systems has made the need for automatic Question Generation greater than ever [5]. Generating questions raises several research issues such as fact selection [6], verbalization, distractor choice and difficulty assessment. These tasks have received a lot of attention in the past few years, the problems related thematic question generation and thematic distractor generation remain largely unaddressed.