Automated Assessment - An Application in Authentic Learning Using Bloom's Taxonomy

In any teaching and learning system assessment is an integral part. In any conventional evaluation process assessment is performed by human instructors or Subject Matter Experts (SMEs). In context to Industry 4.0 Authentic Learning plays a significant role, as well it pursues important role in academia. In case of automated assessment, the system requires classified items of various level of cognitive domain so that thorough assessment can be performed. In the proposed methodology, the authors use pre-classified subjective type items from the Item Bank. Each item belongs to a particular level of cognitive domain with a weighted value assigned to it, these weights related to individual items are predefined by the instructors/SME based on level of mastery in that subject domain along with appropriate level of cognitive domain. In this research, the authors proposed a system where the subjective type of assessment is done based on the pre-classified items and the same using various Deep Learning (DL) and Machine Learning (ML) techniques. Randomly chosen pre-classified items from the Item Bank are selected, which are evenly distributed from the various levels of the cognitive domain of Bloom's Taxonomy. Each obtained response is traversed through a pre-processing module. To extract the features, tokenization is performed on the obtained pre-processed response. The extracted features will be presented in the numerical format and is the input for the regression module. The regression module provides classified responses for each item in the given evaluation. A training set (validation set) which is already available for the evaluation process will compare with the classified responses. Using the above-mentioned workflow based on the metrics like accuracy the performance of a specific learner is measured.