The role of labels in the automatic assessment of graph-based diagrams

The ability to draw diagrams in free-form is rarely found in e-assessment systems. This paper examines one crucial area which needs to be well understood if automatic marking of diagrams is to be feasible: the analysis of labels. Graph-based diagrams include geometrical elements of differing shapes connected by lines to carry the semantics of the domain being modelled. Some diagrams also use the relative location of elements to express information. Labels play a central role in providing meaning both as identifiers, distinguishing between geometric elements, and the values of attributes of elements. Such information is central to the process of comparing two diagrams when marking student-drawn diagrams. Student-drawn diagrams likely to contain errors – they are imprecise. The aim, therefore, is, in the presence of imprecision, to match a student diagram with a model solution and award a mark which is close to that which an expert human would give. This paper explores the specific problem of determining the similarity of labels in diagrams which has been successfully employed in marking student diagrams in a formative environment