SIFVT: A Semantically Inclined Framework for Financial Video Tagging

This paper focuses on the application of metadata and topic modelling to the dataset, which is subsequently subjected to classification via CNN classifiers. This is done by incorporating information from authoritative sources like e-books, which leads to the creation of ontologies that add to the model’s knowledge base. Fuzzy C means clustering plays a pivotal role in grouping entities within the lateral subspace, bridging the gap between those inherent in the dataset and those sourced from the expansive landscape of the worldwide web. The incorporation of metrics such as Adaptive Pointwise Mutual Information (APMI) and Normalised Pointwise Mutual Information (NPMI) serves to amplify the semantic relevance of intermediate terms, resulting in enhanced semantic permeability within the framework. The strategic use of Moth Flame Optimization makes the computation of solution sets even more efficient, which improves the quality of recommended tags as a whole. The proposed model demonstrates its effectiveness with a low False Discovery Rate (FDR) of 0.06, alongside, it achieves the highest mean values across pivotal metrics, including precision (94.45), recall (95.18), accuracy (94.81), and F-measure (94.81). In the context of recommending financial videos, this model emerges as a leading contender, showcasing superior performance compared to established baseline models.