Method of Automatic Construction of Training Exercises for Intelligent Tutoring Systems

The article proposes a method of constructing training exercises by teacher request to balance the concepts being studied and the problem difficulty levels. The method is based on ranking learning problems according to their relevance to the request, taking into account previous problems in the exercise and alternating problems and learning objectives within problem series. Using a problem bank, the key characteristics of the generated problems are identified. The experimental results showed that the method satisfied the stated requirements; removing any part of the method led to a deterioration in the generated problems. The proposed method significantly reduces teachers’ workload when using large learning problem banks.