This study presents MOCS-BERT, a novel extractive text summarization framework that effectively integrates multi-objective Cuckoo Search Optimization with Sentence-BERT embeddings to generate semantically coherent and readable summaries. Evaluated on the full CNN/DailyMail test set comprising 11,490 documents, the proposed model demonstrates statistically significant superiority over three established metaheuristic algorithms namely Bat Algorithm (BA), Flower Pollination Algorithm (FPA), and Firefly Algorithm (FA) as confirmed by the Wilcoxon signed-rank test (p = 0.0432 and p = 0.0010, respectively). Although it does not show statistical significance against Flower Pollination Algorithm (p = 0.0990) among the three baseline metaheuristic algorithms evaluated (BA, FPA, FA), MOCS-BERT consistently achieves the highest ROUGE-L F1 score (0.1895), underscoring its exceptional ability to preserve narrative coherence and logical structure in generated summaries. Furthermore, the model produces highly readable output, with a Flesch Reading Ease of 66.9 and low grade-level indices, making it accessible to a broad audience. These results validate that the integration of deep semantic representations with a carefully designed multi-objective fitness function balancing semantic relevance, non-redundancy, and readability yields a robust, scalable summarization system with balanced performance across semantic coherence, redundancy control, and readability metrics. The research not only progresses the methodological boundaries of metaheuristic-based text summarization but also provides practical utility in real-world applications, including news aggregation, legal document analysis, and emergency response assistance.. Future work will focus on human evaluation, domain-specific adaptation, and automated hyperparameter tuning to further enhance performance and generalizability.
Annisa et al. (Sun,) studied this question.