Automatic question generation from text has been used and adapted to online and self-directed learning platforms. We incorporate methods into the automatic question generation process that are designed to improve question quality by aligning them to the specified pedagogical goals and to a learner's model. This is achieved by extracting, ranking and filtering relevant sentences in the given learning document as well as the questions automatically generated by their semantic associations to the learner model and instructor goals. We propose evaluation techniques for assessing the quality of the questions generated using both human and automatic evaluation.