Word Embeddings and Convolutional Neural Network for Arabic Sentiment Classification

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 2418–2427, Osaka, Japan, December 11-17 2016. Word Embeddings and Convolutional Neural Network for Arabic Sentiment Classification Abdelghani Dahou, Shengwu Xiong, Junwei Zhou*, Mohamed Houcine Haddoud and Pengfei Duan Hubei Key Laboratory of Transportation Internet of Things, School of Computer Science and Technology, Wuhan University of Technology Wuhan 430070, China dahou@whut.edu.cn xiongsw@whut.edu.cn junweizhou@msn.com haddoud.medhoucine@gmail.com duanpf@whut.edu.cn Abstract With the development and the advancement of social networks, forums, blogs and online sales, a growing number of Arabs are expressing their opinions on the web. In this paper, a scheme of Arabic sentiment classification, which evaluates and detects the sentiment polarity from Arabic reviews and Arabic social media, is studied. We investigated in several architectures to build a quality neural word embeddings using a 3.4 billion words corpus from a collected 10 billion words web-crawled corpus. Moreover, a convolutional neural network trained on top of pre- trained Arabic word embeddings is used for sentiment classification to evaluate the quality of these word embeddings. The simulation results show that the proposed scheme outperforms the existed methods on 4 out of 5 balanced and unbalanced datasets. 1 Introduction A growing number of people get used to give their opinions on social network websites, forums, video sharing websites, blogs and e-commerce websites, leading to a most rising research fields caused by the important opinionated web contents. The opinions could be used for many applications such as consumer modeling, sales prediction, opinion survey or user intent understanding. Sentiment analysis which is used to identify, extract and classify subjective information in the opinions, has attracted a lot of attention. Sentiment analysis can be divided into several levels: document level (Yessenalina et al., 2010), sentence level (Farra et al., 2010), word/term level (Engonopoulos et al., 2011) or aspect level (Chifu et al., 2015). Currently, sentiment analysis is commonly used for English, while sentiment analysis on the Arabic language is still recognized at its early stages (Nabil et al., 2015; ElSahar and El-Beltagy, 2015), since sentiment analysis on Arabic is considered as a more challenging work. Firstly, Arabic has a very complex morphology and structure. Inflectional and derivational nature of Arabic language makes the monophonically analysis on Arabic more difficult (Hammo et al., 2002). Secondly, Arabic Internet users mostly use dialectal Arabic rather than Modern Standard Arabic (MSA), while MSA is the formal written language but dialectal Arabic is used in informal daily communication. Moreover, dialectal Arabic is not included in education systems or standardized (Habash, 2010). The diversity of different writings and the language cultural creates more challenges to learn and build language models for representations. Nowadays, more than 267 million people speak Arabic as the first language, and more than 250 million as the second language covering 58 countries1. There are around 168.1 million Arabic Internet users with a user growth rate of 6,592.5% (November 2015 by Internetworldstats2) , making research of sentiment analysis on the Arabic language important. In this paper, we try to …

Content truncated to 500 words