CLASSIFICATION OF ARABIC TEXTS USING DEEP LEARNING

---

Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Erciyes Üniversitesi, Fen Bilimleri Enstitüsü, Türkiye

Tezin Onay Tarihi: 2018

Tezin Dili: İngilizce

Öğrenci: BILAL SHAKIR FARAJ ALKHASAWE

Asıl Danışman (Eş Danışmanlı Tezler İçin): Fehim Köylü

Özet:

The increasing amount of Internet users and the consequent increase of online user reviews, expressing their opinions, has resulted in huge and raw data. This data includes important but unprocessed feedback about products and services. This leads to a growing interest on dealing with the analysis of this type of text data. This area of research is typically called sentiment analysis or opinion mining. In the literature different classifiers are proposed for text sentiment analysis. First raw text data is preprocessed and after that a learning algorithm is used. These algorithms are applied for English language very well. Unfortunately, Arabic is more morphological and wide used language which has different accents used in the world. Support Vector Machines, Decision Trees, and Naïve Bayes algorithms are used for classifying Arabic text in different studies. In this paper, convolutional neural networks (CNN) models are used to classify the Arabic text by using deep learning methods. LABR book review dataset is used for benchmark tests. The results are compared with Support Vector Machines and Naïve Bayes algorithms. In experiments, it is shown that some results of CNN model are better than SVM and NB classification results. Keywords: Text classification, Arabic text mining, sentiment analysis, convolutional neural networks, deep learning.