This repository contains the methods and apporaches used during our thesis work for the short answer scoring problem. The thesis work is divided in three parts. The first part invetigate the feasability of the machine learning approach for short answer scoring using different set of features. The second approach borrow a rule-based method where a set of rules is builded and the student answer must match a suffisant numbers of rules in order to get a credit. The final method combines the two aforementioned approaches in order to make a better model by using a self-learning algorithm augmented by a rule-based system.