A decomposition-based multi-objective optimization approach for extractive multi-document text summarization

Abstract

Currently, due to the overflow of textual information on the Internet, automatic text summarization methods are becoming increasingly important in many fields of knowledge. Extractive multi-document text summarization approaches are intended to automatically generate summaries from a document collection, covering the main content and avoiding redundant information. These approaches can be addressed through optimization techniques. In the scientific literature, most of them are single-objective optimization approaches, but recently multi-objective approaches have been developed and they have improved the single-objective existing results. In addition, in the field of multi-objective optimization, decomposition-based approaches are being successfully applied increasingly. For this reason, a Multi-Objective Artificial Bee Colony algorithm based on Decomposition (MOABC/D) is proposed to solve the extractive multi-document text summarization problem. An asynchronous parallel design of MOABC/D algorithm has been implemented in order to take advantage of multi-core architectures. Experiments have been carried out with Document Understanding Conferences (DUC) datasets, and the results have been evaluated with Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metrics. The obtained results have improved the existing ones in the scientific literature for ROUGE-1, ROUGE-2, and ROUGE-L scores, also reporting a very good speedup.