Conference article

Integrating an adaptive Intelligent Tutoring System (ITS) in real-life school contexts requires coverage of the official curricula, which necessitates a broad range and number of activities to practice the official set of language phenomena. In the context of developing an adaptive ITS for English as a Foreign Language, we propose a method to automatically derive rich activity models from ordinary exercise specifications. The method identifies the language means being covered from the curriculum by processing the language used in the exercise and exemplary answers. The analysis serves two purposes: First, it informs material developers about the extent to which the materials appropriately cover the language means to be practiced according to the curriculum. Second, it helps establish a direct link between rich activity and learner models, as needed for adaptively sequencing activities. The approach includes (1) an NLP-based information extraction module annotating language means using a pedagogically-informed categorization, and (2) a tool to generate activity models offering information on the language properties of each activity in quantitative, qualitative, specific or aggregated terms. We exemplify the benefits of the method proposed in the design of materials for an ITS for language learning used in school.

activity models, automatic annotation of language constructs, evaluation of curriculum coverage in EFL, adaptive sequencing