Effectual LDA Approach for Optimal Topic Modelling from Social Network Data

Methods like document clustering & topic modelling in OSNs allow for the classification, annotation, and analysis of massive volumes of user-generated content. Using these methods, you may finally comprehend the material. Among the various approaches that have been developed throughout time are text mining and clustering techniques. Other techniques, such as hidden subject models & neural embedding approaches, have also been developed. Prior Inference for the Use of Latent Dirichlet Allocation is included as a comparison to our other topic models. Given the many accessible methods of assessment, we review the research to determine which extrinsic measures will provide the best results and then provide guidance on how to use them. Furthermore, we demonstrate the efficacy of the methods on data sets consisting of documents of varied lengths. Our study found that, when using appropriate extrinsic evaluation measures, the best results were obtained across all data sets by using clustering techniques to neural embedding versions of features. Using tf-idf weights & embedding distance measurements, we also provide a top-words based approach to make sense of the clusters. This strategy is predicated on the notion that a sentence's most important words are also the ones that occur most often in it.