Enhance Arabic Clustering using an Educated Text Stemmer with WordNet as Extraction and Water Cycle Cluster 

Abstract

Natural language processing represents human language in computational technique, which is to achieve the extraction of important words. The verbs and nouns found in the Arabic language are significantly pertinent in the process of differentiating each class label available for the purpose of machine learning, specifically in 'Arabic Clustering'. This paper implemented the extraction of verbs and nouns sourced from the Qur'an and text clustering for further evaluation by using two datasets. The limitations of conventional clusters were identified, such as k-means clustering on the initial centroids. Therefore, the current work incorporated a novel clustering optimisation technique known as the water cycle algorithm; when combined with k-means, the algorithm would select the optimal initial centroids. Consequently, the experiments revealed the proposed extraction technique to outperform other extraction methods when using an actual Qur’an dataset. The use of Arabic clustering in assessing the proposed water cycle algorithm in combination with k-means as a clustering further depicted superior performance in harmony search, k-means, and the water cycle. By using the Text REtrieval Conference (TREC) 2001 dataset, the proposed water cycle algorithm in combination with k-means yielded the best score of 0.791%.