Abstract:

Modern teaching, both performed by autonomous education technologies and teachers during classroom events benefit from high-quality student models. Adjusting teaching considering the progress and needs of learners requires a meaningful model based on observations of learning activities in relation to knowledge about processed teaching content.

Current student models, such as Knowledge Tracing Models (KTM) widely-used in Intelligent Tutoring Systems, divide teaching content into Learning Objectives also named Skills describing the main emphasis of teaching goals. They measure learning outcome and probability of teaching success by observing learner’s results during assessments. A main challenge of this approach is to identify relevant Learning Objectives from teaching material to find appropriate Skills to use in Knowledge Tracing Models.

We describe our approach towards a tool chain, that automatically identifies relevant Learning Objectives from teaching material and transfer these to Skills usable in KTM. To lay out a foundation on how to identify Learning Objectives in lecture content, we introduce Generative Learning Theory based Business-Process-Models describing the generative process of preparing teaching materials. These generative processes models provide the fundamentals needed to use Natural Language Processing (NLP), more specifically to build Structural Topic Models (STM), to pull out relevant topics and relations between topics from teaching materials. We introduce a mathematical definition of information based on related data, which allows us to extend the results of these Topic Models and synthesize Topic-Maps. These Topic-Maps, based on connected Learning Objectives and significant teaching material, allow a transfer of STM inferred results into semantic KTM Skills related to appropriate teaching content.

To support our approach, we demonstrate the Structural Topic Model inference of Learning Objectives on lecture material from our university. We synthesize a Topic-Map and infer semantic KTM Skills connected to relevant teaching content. We accompany the results of this demonstration with first results of our interviews of learners and teachers which manually identify Learning Objectives from their attended and prepared lecture documents.

Keywords:

Knowledge tracing model, structural topic model, learning objectives inference, business process models.