Adaptive Filtering of Comment Spam in Multi-user Forums and Blogs

The influence of web-based user-interaction platforms, like forums, wikis and blogs, has extended its reach into the business sphere, where comments about products and companies can affect corporate values. Thus, guaranteeing the authenticity of the published data has become very important. In fact, these platforms have quickly become the target of attacks aiming at injecting false comments. This phenomenon is worrisome only when implemented by automated tools, which are able to massively influence the average tenor of comments. The research activity illustrated in this paper aims to devise a method to detect automatically-generated comments and filter them out. The proposed solution is completely server-based, for enhanced compatibility and user-friendliness. The core component leverages the flexibility of logic programming for building the knowledge base in a way that allows continuous, mostly unsupervised, learning of the rules used to classify comments for determining whether a comment is acceptable or not.