Keyphrase extraction from Arabic scientific articles

Keyphrases are very important tools for summarizing, clustering, indexing and searching documents. Many academic journals request from article authors a list of keyphrases summarizing their research articles. Despite the importance of keyphrases, unfortunately only a few of published Arabic articles contain them. Many algorithms and systems have been suggested and applied by automatically extracting keyphrases for many languages. In contrast to this rich literature, only a few articles have been written for the Arabic language. In this thesis, an attempt will be made to extract keyphrases from Arabic articles, by making use of two methods; the first method uses a specialized stemming approach for extracting keyphrases. The second method splits the articles with respect to their main sections and determines the importance of the phrases in each section. In this research a keyphrase extraction corpora for the Arabic language will be built, a new morphological processing strategy especially for keyphrase extraction will be implemented and this algorithm will be compared with two state-of-the-art algorithms, namely Kea and KP-Miner. The proposed morphological processing algorithm achieves superior results compared to these algorithms.