Extractive single document summarization using multi-objective optimization: Exploring self-organized differential evolution, grey wolf optimizer and water cycle algorithm

摘要

Text summarization techniques become paramount in extracting relevant information from large databa-ses. Current paper attempts to build some extractive single document text summarization (ESDS) systems using multi-objective optimization (MOO) frameworks. Three techniques are proposed: (1) first is an integration of self-organizing map (SOM) and multi-objective differential evolution (MODE) (named as ESDS_SMODE) (2) second is based on multi-objective grey wolf optimizer (ESDS_MGWO) and (3) third is based on multi-objective water cycle algorithm (ESDS_MWCA). The sentences present in the document are first clustered utilizing the concept of multi-objective clustering. Two objective functions measuring compactness and separation of the sentence clusters in two different ways are optimized simultaneously using MOO framework. The proposed approach is able to automatically detect the number of sentence clusters present in a document and then representative sentences are selected from different clusters using some sentence scoring features to generate the summary. The experiments were conducted on two benchmark datasets, DUC2001, and DUC2002, and the obtained results are compared with various state-of-the-art techniques using ROUGE measures. Results illustrate the superiority of our approach in comparison to state-of-the-art techniques in terms of ROUGE−2 score for both datasets. Code of the developed approach ESDS_SMODE is available online at https://drive.google.com/open?id=1WagTeIDLgphttPrKHpnF_eO7QHWJxXxK.