In order to boost both efficiency and effectiveness of meetings, we propose a novel Meeting Warming-up system to detect common interests and conflicts among participants before a meeting. The basic idea of the proposed approach is: firstly, modelling user preference by extending the attribute concept tree with additional relations both in and across attributes; secondly, determining common interests and conflicts through preference propagation and merging; thirdly, visualizing the detected results via a group preference graph. As a result, each participant can intuitively understand the group's opinions as a whole and warm up for discussions around potential outcomes. In particular, the meeting may have an easy and friendly start with commonly agreed outcomes, the commonly disagreed items may be ignored to save time, and participants may be mentally prepared to discuss the possible conflicts carefully and sufficiently. The experimental results showed that our approach is feasible.