A Region-based Approach to the Automated Marking of Short Textual Answers

Automated marking of short textual answers is a challenging task due to the difficulties involved in accurately “understanding” natural language text. However, certain purpose-built Natural Language Processing (NLP) techniques can be used for this purpose. This paper describes an NLP-based approach to automated assessment that extends an earlier approach [1] to enable the automated marking of longer answers as well as answers that are partially correct. In the extended approach, the original Question Answer Language (QAL) is augmented to support the definition of regions of text that are expected to appear in a student's answer. In order to explain the extensions to QAL, we present worked examples based on real exam questions. The system's ability to accurately mark longer answer texts is shown to be on a par with that of existing state-of-the-art short-answer marking systems which are not capable of marking such longer texts.