Extractive Single Document Summarization Using an Archive-Based Micro Genetic-2

Research on automatic text summarization is growing with the rapid growth of online information services. In the current study, Extractive Single Document Summarization (ESDS) is formulated as a multi-objective optimization problem, and an Archive-based Micro Genetic-2 Algorithm (AMGA2) is employed to solve this problem. In this approach, each summary is represented in a binary form. The initial population is created using the clustering and Latin Hypercube (LH) sampling, resulting in a good overall unbiased population distribution in the genotypic space. Two objective functions were formed considering content coverage and informativeness as objective function 1, and anti-redundancy as objective function 2. The summary length limit is considered as the constraint. For generating offspring, an archive-based mating pool and modified mutation operator of differential evaluation are used. Finally, from the optimal set of solutions, the best solution is selected to generate the summary. For evaluation, DUC-2001 and DUC-2002 datasets are utilized, and the obtained results are compared with various existing methods using ROUGE measures. The obtained result illustrates the superiority of our approach in terms of convergence rate and ROUGE scores.