Semantically enriched cloze exercise — Technical concept and prototype

Abstract

In our work we describe an approach for automated and approximate evaluation of character based answers in computer-based tests. Cloze exercises (‘fill-in-the-blank’) are error-prone to a variety of potential error sources and are subject to syntactic and semantic ambiguities. Our approach combines string similarity measurements with semantic word analysis to render this type of computer-based exercise more robust and user-friendly. We present a working prototype for a web-based e-learning system supporting spelling correction (syntax level) and an extension to synonyms (semantic level). A community-based approach using OpenThesuarus.org is proposed to gather and evolve missing synonyms for given answers of cloze exercises.