Web image retrieval is semantically inclined, and knowledge driven is the need of the hour due to increase in the multimedia contents on the world wide web. In this paper, web image retrieval framework, which is semantically driven knowledge centric Web 3.0 complaint has been proposed. The proposed framework MOWIRframework is a Metadata Centered Ontology Controlled Web Image Recommendation Scheme for Botanical and Horticultural Domains. In this framework Ontology Modeling and Generation is coupled with differential Classification using the LSTM and Logistic Regression classifiers at varied levels. The Ontology Alignment tasks ensures a fair amount of auxiliary knowledge is anchored with a high degree of relevance to the domain and the Metadata Generation exponentially increases the knowledge from global Web to the localized framework for recommendation reducing the cognitive knowledge gap. This framework strategically integrates a feature-controlled Machine Learning Classifier and a Deep Learning Classifier at two distinct instances to enhance variational diversity in learning. An array of semantic similarity computation measures with a differential threshold and deviance criterion are employed in the framework for increasing the strength of relevance computation. The proposed MOWIR achieves an overall Precision of 93.11% with an average F-Measure is 95.24% with a very low False Discovery Rate (FDR) of 0.07 which outperforms the baseline approaches

Semantically driven, LSTM, Logistic Regression, Text classification.