On the Creation of Classifiers to Support Assessment of E-Portfolios

Text-classifiers are among the important services required of contemporary AI systems. Based on a trained classifier, one can perform relevant tasks such as highlighting relevant passages of a text, analyzing what is talked about or encouraging students to write texts which cover important concepts. This form of analysis is at the core of the AISOP project, AI-supported Observation of e-portfolios: Through the development of classifiers and the visualization of their results, the project aims at supporting e-Portfolio assessment in the university context. We report our investigations on the creation process of classifiers, on how their quality can be evaluated and enhanced, and on the domain specificities we have met.