Learner satisfaction is the degree of consistency between learners' perception and expectation of learning experience. In massive online open courses, analyzing the influencing factors of learner satisfaction is of great significance to improving the quality of course development and learning experience. Taking the open course reviews as data source, the paper adopted topic sentiment analysis and intermediary hierarchical linear modelling to analyze the impact of different student and course level features on learner satisfaction. The data analysis shows that the schedule, workload and completion status, as well as the video, instructor, content and evaluation topics play significant roles in explaining learner satisfaction; However, perceived difficulty, structure and interaction are not related to learner satisfaction; Meanwhile, sentiment-mediated analysis found that, teachers and evaluation topics have significant mediating effect on schedule; Video topic has a significant mediating effect on workload. Based on the above analysis, enlightenment to curriculum designers and developers are also provided.