In order to deal with the rapid and diversifi ed development of spam messages

the governance means of operators towards spam messages need to advance with the times

need to cover more comprehensive and judge more accurate. In view of the objectives

we proposed an identifi cation and fi ltering method based on prior model for short text spam messages. This method analyze short text from three aspects

respectively text feature layer

keyword pattern layer and content feature layer

then ultimately give a final judgment result using comprehensive analysis and determination layer with previous training process. We can not only ensure the precision of spam messages recognition

but also greatly improve the recall of spam message recognition. We give a better solution to the current spam messages recognition problem

which may save a lot of manpower for the operators.