Document Novelty Detection is a concept learning problem wherein the system gains its knowledge only from the positive documents under a concept and with that limited knowledge it attempts to detect the negative cases. This work focuses on learning author style as a concept from the given set of documents, particularly e-mails. Since author attribution for smaller texts such as e-mails is more complex compared to larger documents, the techniques originally used for the large documents prove inefficient for smaller texts. The main goal of this work is to address this shortcoming of existing algorithms in detecting aberration in author style.
A graph model based technique for feature set extraction from small documents has been proposed and evaluated. Also two probability based text representation schemes have been developed that could best represent a text document to an underlying one-class SVM classifier. The proposed models have been compared and evaluated against the public Enron e-mail dataset. Applying graph based feature set extraction technique in combination with the inclusive compound probability based text representation has proved to be very efficient and hence we have extensively evaluated the effect of all controlling parameters to arrive at the optimal values.
