Automatic summarization of the Arabic documents using NMF: A preliminary study

The exponential growth of the Internet produces a huge amount of documents online. Finding the desired documents from amongst these huge resources is a difficult task. This problem is known as “Information Overloading”. Automatic Text Summarization techniques (ATS) try to solve this problem by extracting the essential sentences that cover most of the main issues in the document. So the user will spend less time and effort to identify the main ideas of the document. Research in this field in the Arabic language is relatively new compared with the available research in English. This paper presents a preliminary study that investigates the effectiveness of using Non negative Matrix Factorization (NMF) algorithm to summarize the Arabic documents. The researcher of the present study has built an Arabic corpus of 150 documents manually and conducted extensive experiments by using different sentences scoring algorithms and term weighting schemes. The performance of the proposed algorithm has been measured, and the extensive experiments have shown that the NMF algorithm yields promising results.