Classification of extremist text on the web using sentiment analysis approach

The high volume of extremist materials online makes manual classification impractical. However, there is a need for automated classification techniques. One set of extremist web pages obtained by the TENE Web-crawler was initially subjected to manual classification. A sentiment-based classification model was then developed to automate the classification of such extremist Websites. The classification model measures how well the pages could be automatically matched against their appropriate classes. The method also identifies particular data items that differ in manual classification from their automated classification. The results from our method showed that overall web pages were correctly matched against the manual classification with a 93% success rate. In addition, a feature selection algorithm was able to reduce the original 26-feature set by one feature to attain a better overall performance of 94% in classifying the Web data.