| 摘要 |
<p>In recent years, spam has become a major problem with a huge impact on society. Automatically filtering such messages poses a special challenge in terms of text categorization, in which the key characteristic is that the filters are fighting against an active opponent which is always attempting to evade filtering techniques. The present invention relates to a method for automatically filtering spam based on the principle of the most simple description supported by confidential elements. The MDL-CF method has a number of suitable characteristics: it consumes limited processing resources, it is easy to implement, very quick to classify new messages and it is very easy to train and teach. It classifies spam effectively and efficiently, achieving better results than established methods and other filters previously published in the literature. Another interesting point is that the method proposed can be extended and applied to address problems in different areas involving text classification and categorization, such as: web spamming, blog spamming, social network spamming, mobile spamming, and any other type of text-based spam.</p> |
