Automatic multi-documents text summarization by a large-scale sparse multi-objective optimization algorithm

Abstract Due to the exponential overflow of textual information in various fields knowledge and on internet, it is very challenging extract important or generate a summary from some multi-document collection specific field. With such gigantic amount content, human text summarization becomes impractical since expensive consumes lot time effort. So, developing automatic (ATS) systems becoming increasingly essential . ATS approaches are either extractive abstractive. The approach simpler faster than abstractive approach. This work proposes an system that aims small subset sentences large text. First, whole preprocessed by applying natural language processing techniques as segmentation, words tokenization, removal stop-words, stemming provide structured representation original document collection. Based this representation, problem formulated multi-objective optimization (MOO) optimizes extracted maintain coverage main content while avoiding redundant information. Secondly, evolutionary sparse algorithm developed solve large-scale MOO. output set non-dominated summaries (Pareto front). A novel criterion proposed select target Pareto front. has been examined using (DUC) datasets, have evaluated (ROUGE) metrics compared with literature.