This paper discusses a model for structuring resources for automatic assessment in scientific education by means of textual descriptions. This study aims to support instructors in extending teaching strategies and expanding formative assessment in virtual communities of practice. The strategy to achieve these goals involves the implementation of a system for adaptively providing the students with instructional activities based on Natural Language analysis and ontological knowledge bases. The model is presented, and the feasibility of its use is discussed taking into account the results collected from the experimentation with instructors.