Nowadays there are numerous user-generated restaurant reviews available on the Internet, of which they are considered valuable resources for decision making to customers. In reality, not every reviews available online are helpful to users, so the need for filtering unqualified reviews is realized. There have been several studies on spam review detection that attempt to detect unqualified reviews using n-gram linguistic features. They attempted to classify reviews written in English into two categories: filtered and unfiltered reviews. In our paper, we are aware that filtered reviews can be categorized further into advertisements, complaints, and non-review opinions. Each type of filtered reviews requires different actions be taken on them. For instance, advertisements should be deleted while the others should be hidden. This paper proposes a framework for unqualified restaurant reviews detection using SVM. The data set of restaurant reviews is provided by a web portal for restaurants in Thailand. There are three main processes in our framework: (1) filtering obvious spam, (2) data pre-processing, and (3) classification using SVM. We find that the procedures in our framework can improve classification accuracy compared with standard frameworks.