A survey of text summarization and Headline Generation methods in Arabic

HEADLINE GENERATION IS CONSIDERED ONE OF THE CHALLENGING TASKS IN TEXT SUMMARIZATION. THE INCREASING NUMBER OF SUCCESSFUL MODELS IN THE TEXT SUMMARIZATION FIELD HAS SHOWN THE ABILITY TO AUTOMATICALLY GENERATE INFORMATIVE HEADLINES FOR ARTICLES INSTEAD OF HAVING CATCHY, MISLEADING HEADLINES WRITTEN BY JOURNALISTS. HOWEVER, MOST OF THE CURRENT WORK IN HEADLINE GENERATION FOCUSES ON GENERATING ENGLISH HEADLINES ONLY, WITH VERY FEW PAPERS PROPOSING ARABIC HEADLINE GENERATION AND USING SMALL-SIZED DATASETS COMPARED TO ENGLISH DATASETS. THIS PAPER DESCRIBES A LITERATURE SURVEY OF TEXT SUMMARIZATION AND HEADLINE GENERATION METHODS IN ARABIC. SUMMARIZATION SYSTEMS, IN GENERAL, CAN BE MONOLINGUAL OR MULTILINGUAL SYSTEMS WHERE SUMMARIZATION METHODS CAN BE CLASSIFIED ACCORDING TO SEVERAL CATEGORIES, SUCH AS LINGUISTIC OR STATISTICAL AND EXTRACTIVE OR ABSTRACTIVE. THEREFORE, WE WILL HIGHLIGHT THE AVAILABLE RESEARCH AND APPROACHES USED IN ARABIC. WE WILL DISCUSS THE APPROACH ARCHITECTURE, USED DATASETS, EVALUATION METRICS, AND THE RESULTS FOR EACH APPROACH. FINALLY, IN CONCLUSION, WE WILL DISCUSS THE POSSIBLE FUTURE WORK FOR HEADLINE GENERATION IN ARABIC.