Comment-based summarization using a multi-objective cuckoo search optimization approach

In the digital era, user-generated content in the form of comments and discussions on online platforms has grown exponentially. This proliferation underscores the increasing significance of effective automatic methods for comment-based summarization. The goal of these methods is to generate a summary that covers the most important information from both source documents and user-generated content. In this work, the Comment-based Summarization problem has been addressed with a Multi-Objective Cuckoo Search algorithm (CSMOCS). The criteria of content coverage, comment relevance, and redundancy reduction have been formulated as the objective functions to be optimized. New problem-aware evolutionary operators have been proposed. Experiments have been conducted using the Reader-Aware Multi-Document Summarization (RA-MDS) dataset, and model performance has been evaluated with standard Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metrics. In comparison to other eighteen methods from the literature, the proposed CSMOCS approach reports average percentage improvements of 25.08 % for ROUGE-1, 33.21 % for ROUGE-2, 20.75 % for ROUGE-L, and 29.03 % for ROUGE-SU4 scores. These results denote the ability of the proposal to automatically generate high-quality summaries taking into account user comments. • Comment-based summarization problem is addressed from a multi-objective viewpoint. • First multi-objective swarm-intelligence algorithm for solving the presented problem. • Problem-specific algorithmic strategies are devised for mutation and repair operators. • An exhaustive comparative study with other eighteen methods and statistical metrics. • Significant improvements in detecting unigrams and harder constructions like bigrams.