Reflection has been shown to be a effective way for students to learn, and a lot of instructors have adopted it. Although it is time consuming for instructors to read those reflections, reading them is valuable to instructors because they are a rich source of ideas on improving teaching. Therefore, there is demand for systems that allow instructors to quickly grasp actionables raised by students. We propose a novel way to extract actionable insights that instructors can use to improve their teahcing from a massive text data of student reflections. They can see what topics their students are talking about and how they feel about those topics through clustering of reflecitons and their summaries.