A Vector Space Model based spam SMS filter

摘要

Along with the popularity of telecommunication and mobile phone, short message (SMS) enters almost every human life. Meanwhile, each mobile phone client suffers from the harass of spam SMS. As the SMS service provider who is in charge of all industry SMS in east China, Dahan Tricom Corporation always invest much in anti spam SMS research. Recent years, we upgrade our anti-spam filter to semantic level. The core technology is described in this paper. Unlike other anti spam filter, such as anti spam emails, the anti spam SMS filter must face many difficulties oriented by SMS itself. Since SMS contains only 70 Chinese characters or 140 English letters at the most, it is always lack of semantic information. Also, vocal expressions often appear in SMS. In addition, the industry SMS often contains proper terms related to specific industry field. The anti spam SMS filter in this paper first leverages Vector Space Model (VSM) as its foundation technologies. Then, many modifications are made in the process in VSM method to address the difficulties of spam SMS filtering issue. Finally, the experiment result turns out to be acceptable in our commercial production environment.