Similarity Measures for Classical Arabic Poetry Ranking

Abstract

Arabic language is a very complicated language. The Classical Arabic poetry has difficult words in their meanings and grammar. Although these poetries use a very complex Arabic morphology it is not a tedious task to classify them. This approach is presenting a method to rank the poetries of a certain age in the Arabs history to their main ranks (like: Ritha'a, Hija'a, Ghazal…etc,) using the algorithm of N-grams statistical technique. The whole work is depending on normalizing the poetries without using any stemming process and using the n-gram (bi-grams) for retrieving characters. Retrieval influence tests have been implemented using similarity coefficients: the Dice, the Overlap and the Jaccard coefficient measures. For system evaluation Precision, Recall and F-measure have been calculated. Despite the fact that the Overlap measure seemed to be the best measure the rest gave good results as well.