1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Morphological disambiguation of the Quranic Arabic using classifiers combination
Authors:	Mohammed Nidham Omar Alaani (P66177)
Supervisor:	Nazlia Omar, Assoc. Prof. Dr.
Keywords:	Morphological disambiguation Quranic Arabic Classifiers combination Universiti Kebangsaan Malaysia--Dissertations
Issue Date:	24-Sep-2014
Description:	Arabic is a morphologically rich language, where significant information concerning syntactic units is expressed at the word level, which makes part of speech (POS) tagging a challenge since it involves morphological disambiguation. Morphological disambiguation is the process of assigning POS, tense, number, gender, and other morphological information to each word in a sentence. Arabic has rich morphology, which presents an interesting problem for machine learning and statistical tagging models as the level of ambiguity is high and the potential tag set size is very large. In addition, the ambiguity is higher and the data sparseness is much harder due to the large tagset. Most morphological disambiguation model are either rule-based, stochastic or machine learning. Up to now, there are very few researches conducted on Arabic morphological analysis and disambiguation. This research area has not been studied well and even the existing work only reported modest accuracy. The goal of this thesis is to design and implement efficient and robust morphological disambiguation model for Quranic Arabic corpus (QAC). This thesis also experiments four of combination classifiers framework i.e. Hidden Markov Model (HMM), K-Nearest Neighbour (K-NN), Na've Bayes (NB) and Support Vector Machine (SVM) for handling the problem of morphological disambiguation for Quranic Arabic corpus (QAC). For the evaluation, we particularly compared two methods i.e. direct classification and single -attribute classification. The result of the classifier combination model achieved an overall accuracy of 91.6% and F-measure of 85.80%. Overall, the experimental results show that the combination of single -attribute classifiers improves the classification effectiveness in morphological disambiguation.,Master/Sarjana
Pages:	71
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.