A Biased Random-key Genetic Algorithm for Extractive Single-document Summarisation

Extractive text summarization has been dealt with by several metaheuristics that proved their efficiency. In those works the feasibility of solutions has been mostly guaranteed through some operators, whose role is to check and/or correct infeasible solutions. To reduce the complexity of the task, this works proposes a Biased Random-Key Genetic Algorithm, with a newly-proposed decoder, it is adapted to extractive single-document summarization. We have tested the performances of our approach on two standard datasets, DUC-2001 and DUC-2002, through using the ROUGE-1 and ROUGE-2 metrics. The results are very promising and show that our approach outperforms other reference methods, it came first out of 14 algorithms.