Due to the huge and rapid growth of online data makes search such massive data collections and finding the relevant information a tough task and time consumption. For this reason, research on automatic summarization techniques has received much attention from industry and academia. Unlike English text which has received much attention of the researchers in this field, Arabic text is still lake to such serious investigations. This reason gave the author of this paper, strong motivation to participate in a pushing Arabic language into the concern domain of automatic text summarization researchers by proposing an extractive summarization method. The proposed method generates a summary of an original document based on a linear combination of text features having different structures. Five summarizers (AQBTSS, Gen–Summ, LSA–Summ, Sakhr and Baseline–1) are used in this study as benchmarks. The proposed method and the benchmarks are evaluated using EASC – the Essex Arabic Summaries Corpus. The results showed that the proposed method performs well, based on recall, precision and average scores, more than the five benchmarks. A good performance achieved by the proposed method proved that the focus on those more complicated features, rather than simple ones, could guide to the most important content of any document.
展开▼
