SITSL: Strategic Image Tag Recommendation Framework Encompassing Semantics Oriented Learning and Reasoning

This paper introduces the concept of a strategic structure for Web Image Tagging that is oriented toward semantics and knowledge centric. It encompasses strategies like LDA (Latent Dirichlet Allocation) based topic modelling and ontology generation through the dataset terms and further help in the metadata generation. The generated metadata is classified using autoencoders classifier to yield the classified instances. In the proposed model, the classifier for logistics is applied in the dataset of web images that integrates features through Lin Similarity from the ontologies generated in the perfect global dataset. The LDA also enhances the topics and contextualizes that for classified instances of the dataset. The normalized compression distance with lance William index helps in semantics similarity computation for semantics measure-based reasoning and learning. Shuffle frog algorithm helps in Optimization helps in refining the initial solutions into a much more optimal solution state to produce in the top-tier images. So, to produce in the top-tier tags for the images and overall precision of dash with a measure of dash and an FDR of dash has been achieved by the proposed framework which is the best-in-class model for web image tagging.