Abstract: Question Generation (QG) receives increasing research attention in the NLP community. One QG motivation is to facilitate the preparation of educational reading practice and assessments. While significant advancement of QG techniques was reported, we find current QG techniques are short in terms of controllability and question difficulty for educational applications. This paper reports our studies toward the two issues. First, we report a state-of-the-art examlike QG model by advancing the current best model from 11.96 to 20.19 (in terms of BLEU 4 score). Second, we propose a QG model that allows users to provide keywords for gUIDing QG direction. Human evaluation and case studies are conducted to demonstrate the feasibility of controlling question generation direction.