Structural learning framework for binary short text classification

With the fast-paced prevalence of smartphones, binary short text classification (STC) is becoming a basic and challenging issue, and relevant STC algorithms can be successfully used in spam filtering for short message service (SMS), wechat, microblogging, and so on. In this manuscript, we address the structural feature of SMS documents and propose a structural learning framework, which decomposes the complex binary STC problem according to the SMS document structure, and predicts the final category by combining several sub-predictions. Supported by our index of string-frequence, we also implement some STC domain classifiers. The experimental results show that the performance of two previous STC algorithms can be upgraded by the structural learning framework, and our STC domain classifiers can achieve the state-of-the-art performance on the task of Chinese SMS spam filtering within the structural learning framework.