There exists a requirement to tag content on the web and specifically tweets as the social web has a lot of activity, an uptick in recent years caused by the introduction of a considerably high number of users on the social web and specifically Twitter which attracts a lot of celebrities. This paper proposes the folksonomy-based Twitter tag recommendation framework which encompasses the prominence of building and generating a folksonomy from the Twitter dataset. The FTISI also focuses on enriching the folksonomy using Linked Open Data Cloud (LOD), eBooks, indexes, and several other news repositories. This model extracts the categories from the dataset and applies the Latent Dirichlet Allocation (LDA). The FTISI model fits in the LDA for enriching the categories in the dataset and it is subjected to TF/IDF to derive the informative term set from the historical documents, and the entities derived from the TF/IDF and the LDA and the Dspace tool are utilised in creating the metadata which is in turn classified through the usage of a strong deep learning BiLSTM model. Subsequently, the latent Twitter Semantic Similarity (TSS) using COSIM rank is achieved by encompassing the top 20% of classified metadata and the generated taxonomies. The entities which come out of the lateral semantic similarity computation using COSIM rank and the rest of the 70% of the metadata are subjected to the computation of the SOCPMI under the stochastic optimization algorithm and TSS is also computed using empirically decided thresholds and values and these units are ranked in the rising order of TSS and is sent for review which is conceived as tags. The highest primary evaluation metrics and lowest FDR has been achieved by the FTISI model.