Improving the Cognitive Levels of Automatic Generated Questions using Neuro-Fuzzy Approach in e-Assessment

Assessment is a fundamental activity to realize the objective of a course and to enhance the teaching learning process. Ensuring quality in the question papers for testing the different cognitive skills is important in the test or examination component of any academic or training domain. Bloom's taxonomy, a popular framework has been used to assess the learning skills of students. This paper describes the methodology to auto-generate questions based on bloom's cognitive levels. Various Natural Language Processing (NLP) techniques are incorporated to construct a textgraph using input statements from the web where native intelligence acquired from ConceptNet interrelates the nodes. The proposed work resolves the complexity of categorising autogenerated questions with similar action verbs. It follows a combinatorial formulation of fuzzy logic to generate cognitively fluent questions. A Fuzzy Restricted Boltzman machine coupled with Gaussian Markov based softmax is used in the proposed architecture. Experiments reveal the significance of the proposed system in generating cognitively fluent questions when the same action verb gets interlinked with different cognitive level of auto-generated questions.