Differential evolution cluster-based text summarization methods

• Main Authors: Abuobieda, Albaraa, Salim, Naomie, Binwahlan, Mohammed Salem, Osman, Ahmed Hamza
• Format: Conference or Workshop Item
• Published: 2013
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/50991/; http://dx.doi.org/10.1109/ICCEEE.2013.6633941
• Institution: Universiti Teknologi Malaysia

In this paper, three similarity measures; Normalized Google Distance (NGD), Jaccard and Cosine Similarity measures were employed and tested for textual based clustering problem. A robust evolutionary algorithm called Differential Evolution algorithm was also used to optimize the data clustering process and increase the quality of the generated text summaries. The Recall Oriented Under Gisting Evaluation (ROUGE) was used as an evaluation measure toolkit to assess the quality of the summaries. Experimental results showed that all of our proposed methods outperformed the benchmark methods. More importantly, the Jaccard-similarity based method surpassed all the other proposed methods in this study.