Tweets exchanged over the Internet is an important source of information even if their characteristics make them difficult to analyze (e.g., a maximum of 140 characters). In this paper, we address the problem of analyzing the opinions expressed through tweets for di erent communities. More precisely we are interested in following, over time, what is the opinion that a community can have for a speci c term or expression (e.g., is the opinion of tweets using the term "crisis" remain the same over time for a political party?). Furthermore we are also interested in shared terms or expressions between di erent communities. In this case our goal is to evaluate if the opinion expressed changes a lot between communities. Conducted experiments on tweets for the upcoming French Presidential election show very interesting results.