Abstract
Teaching and instruction use symbolic knowledge, which can be stored in knowledge bases. Knowledge-based educational tools can provide help in many ways for the teaching process. However, what if the knowledge that is necessary is not in symbolic form and cannot be neatly stored in a knowledge base? What if this knowledge has been previously learned by an Artificial Neural Network (ANN) and is embedded sub-symbolically in the weights and biases of that ANN? The instructor can access the input-output structure of the ANN and get responses and predictions, but cannot access the knowledge itself. This is where rule extraction from ANNs finds a yet unexplored research and application field. We have built a knowledge extraction environment for education assistance. In this paper, we discuss our research goal, the functions performed by the environment, and some tests showing the validity of both research and prototype system.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call