TESA: Tagging of Educational Videos Using Semantics Oriented Artificial Intelligence

This paper proposes a tag recommendation model for tagging educational videos powered by semantics-oriented learning and reasoning. The framework is in accordance with the standards of the Web3.0 and applies the CNN classifier into the educational video dataset twice independently using contents and annotations. The entities of the dataset undergo RDF generation from which RDF subject and object are obtained separately. The model then subjects the result to the generation of auxiliary knowledge using Google’s KG API and Wikidata. This is included in the model to improve the number of concepts in the dataset which are distinct and diverse. The metadata generation helps in increasing the amount of auxiliary knowledge added into the dataset which is classified using the Deep Belief Networks (DNBs), a strong deep learning classifier. Semantics-oriented reasoning is, upon inference, through semantic similarity measures like the Second Order Co-occurrence PMI and the Petriatis Index in the model, inducted into the framework. Fire Hawk Optimizer (FHO) is an optimization algorithm that helps in yielding the best-in-class optimal solution set which is subjected to review and comprises the final tags for the educational videos. The framework yields the highest values of primary evaluation metrics as well as a low FDR of 0.04.