Extractive arabic text summarization: a hybrid approach through genetic algorithm-based feature optimization

In this study, a single-document, extractive-based automatic summarization model for Arabic texts has been developed. Within the proposed method, sentences are represented by eight distinct statistical and semantic features: position, length, keyword density, presence of numerical expressions, inter-sentence similarity, identification of named entities, occurrence of Arabic–English mixed terms, and term importance determined by the TF–ISF method. The weights of these features were optimized using the Random Search algorithm, and subsequently, a Genetic Algorithm (GA) was employed to select the most appropriate combinations of sentences. The model was tested on the Essex Arabic Summaries Corpus (EASC) dataset and evaluated using the ROUGE-1, ROUGE-2, and ROUGE-L metrics. The experimental results indicate that the proposed method is capable of generating informative, coherent, and semantically rich summaries.