The text information flow of SMS had carried abundant information resources. In order to find out the hot events behind it, an online sorting algorithm was given for the text information flow of SMS. This method used the cooccurrence frequency of feature words to define its relevance. And the similarity of message texts was defined on the basis of preamble information collection and information frequency. Furthermore, after each time period of clustering, the clustered SMS texts were classified periodically. This algorithm had higher efficiency to find hot events for a large number of short text information retrieval. Moreover, this algorithin reduced the possibility of false detection and missed detection. Based on the comparison of experiments on algorithms between real data sets and SinglePass, the results showed that each index was improved to some degree.