Arabic Texts summarization by Using RST Algorithm

Abstract

The evolution of the World Wide Web led to increase the size of Arabic texts on the Internet, which led to ask many questions about how to behave in this huge and growing number of these texts. And therefore need sure has become more than ever to provide the necessary means for rapid browsing of the texts, in order to enable the user to estimate the degree of relevance of the document to the required information. The search has been subjected to debate automatic summarization for Arabic texts. Because Arabic Language has special features made it one of the languages that need to be carefully to understand the components and the characteristics and difficulties they experience in controlling the formation of letters and syntax sound. Were mentioned how the syntax of the Arabic language by reference to the word roots. Also we mentioned the concept of automatic summarization of the various theories and definitions with examples to the various programs that support automatic summarization of texts, whether in Arabic or English We taken SARA © program as a case study for the programs that support the summary Arabic texts which designed by RDI Egyptian company which specialized in programs that serve Arabic. This research introduces evaluation methods for an Arabic extractive text summarization system by calculating the Recall, precision and F-measure. The system is trainable and uses manually annotated corpus. We have introduced methods for evaluating the summary against other human summaries. Moreover, we used human judgment for system output, and finally we tested the system against a commercial Arabic summarization system. We selected several different Arab texts from Web sites, taking into account the difference in the content of the text, for example, the historical text and text and text scientific news ... etc. The texts were Summarized by SARA © program and also summarized by a human specializing in the Arabic language and were compared between them. We have acquired a good encouraging results prove that the system can be relied upon it.