Mobile Spam Filtering base on BTM Topic Model

At present, Short Message Service (SMS) is widespread in many countries. Many researchers usually use conventional text classifiers to filter SMS sparn. In fact, the actual situation of SMS spam messages isn't consideration by most reseachers. Because the obvious characteristic is the content of SMS spam messages are miscellaneous, shorter and variant. Therefore, traditional classifiers aren't fit to use for SMS spam filtering directly. In this paper, we propose to utilize A Biterm Topic Model(BTM) to identify SMS sparn. The BTM can effectively learn latent semantic features from SMS spam corpus. The experiments in our work show the BTM can learn higher quality of topic features from SMS sparn corpus, and can more effective in the task of SMS spam filtering.