A systematic approach to the automated marking of short-answer questions

Abstract

A number of attempts have been made to automatically mark short-answer questions. The authors of this paper have developed and evaluated a prototype software system that provides a novel approach to automated marking. Capabilities of the recently developed Stanford Parser (such as the ability to produce phrase structure and typed dependency parses of the text of a student's answer) have been utilized in the prototype system. A new notation technique to represent the required syntactical structures has also been developed and is introduced in this paper. A number of real exam questions have been used to evaluate the prototype. The evaluation results are encouraging and the authors intend to extend the system so that it becomes capable of marking longer answers.