There has been an increasing interest in analyzing social network services data. However, detecting social topics in the era of information explosion requires state-of-the-art analytics techniques. The geographic clustering analysis based on social topics across provinces, i.e., states, has rarely been studied. Using the Twitter data collected in the United States (US), we detected the social hot topic by using the ratio of word frequency. Also, we found geographic communities by correlating the time series for a set of topic words across US states. The result of the geographic clustering was visualized using the Google Fusion Table. In conclusion, the ratio of word frequency properly detects social topics or breaking news while suppressing daily tweeted small talks or emotional words such as lol, like, and love. We have also demonstrated that a clustering algorithm based on a social topic can be useful in classifying social communities.