College of Science, Mustansiriyah University discussed a Ph.D. thesis on the effectiveness of an intelligent topic modeling system based on multi-objective optimization for analyzing complex data and providing effective solutions by the postgraduate student, Ms. Rana Faris Najeeb Saeed.

The thesis aimed at automatically detecting hidden topics within large sets of text, this helps organize unstructured data by grouping similar content, extracting valuable insights and reducing dimensions, clearly labeling topics facilitates understanding and interpretation, enhancing decision-making in various fields such as market research, social media analysis and academic research.

The thesis reviewed multi-objective topic modeling and automatic labeling, the topic modeling process includes preprocessing and feature engineering using the SR-LW algorithm, semantic representations using SBERT, dimensionality reduction using UMAP, and topic generation using an improved clustering algorithm using a genetic algorithm, along with parameter optimization for clustering algorithms such as HDBSCAN.

The thesis concluded that the best metrics were achieved using the mini-batch algorithm, with a coherence value of 0.605, diversification of 0.972, and an execution time of 4.762 seconds.