Suggestion Mining: Detecting Suggestions for Improvement in Users' Comments

In the context of the development of a feature-based opi- nion mining system for English, we observed that there is some very interesting information provided by customers but not yet covered by \standard" opinion mining techniques: opinion mining aims at detecting whether comments are positive or negative, but it appears that customers are very often suggesting improvements about what they are reviewing, which is quite dierent from expressing an opinion. This papers proposes to apply Natural Language Processing techniques in order to address this rather new task of extracting automatically such kind of suggestions for improvement from user's comments.