SARAC: A Framework for Automatic Item Generation

Automatic item generation can derive items from certain computational structures (models). This generative approach introduces new features into computer-based assessment and has attracted many researches. However, to construct models for courses in higher education, it usually requires decent programming skills and leads to monolithic systems. To solve this problem, we propose the SARAC (sharable and reusable assessment components) framework. In this framework, item models are stored in XML format. Each item model consists of the skeleton of a group of items and some parameterizable elements. These elements - some of them convey complex knowledge, are produced by distributed components, which we name SARAC. This method localizes programming in components, and therefore makes authoring item model a simple task similar to drag-drop operations.