Normally when developers obtain defects list from users, the development team will decide which defects should be fixed first. The software maintenance plan, which consists of list of defects to be fixed sequentially, is mostly generated using developer experience to prioritize the defects. With the current strategy, the software maintenance plan may not serve well to customer needs. This research proposes a method to prioritizing the software defects to be fixed by focusing on 3 impact factors: severity, priority and the number of users who found the same defects. The proposed method uses natural language process (NLP) to analyze user feedbacks in order to extract defect-related keywords. The ranking process uses analytical hierarchy process (AHP) to prioritize defects to be resolved. An evaluation of the prioritizing method is provided.