Impact of Neural Network Architectures on Arabic Sentiment Analysis
Hamza Chahidi, Hicham Omara, Mohamed Lazaar, Mohammed Al Achhab · 2019
Sentiment Analysis (SA), commonly known as opinion mining, during last couple of years, it becomes the fastest growing research areas in computer science. Conventionally, it helps to automatically detect if a text express is a positive, negative or neutral opinion. It enables us to identify and extract subjective information in a piece of writing, and this leads to gain an overview of wider public opinions or attitudes toward topics, products or services. Many researches have been done in this area, but most of them have focused on English and other Indo-European languages. Insufficient studies have actually accosted Sentiment Analysis in morphologically rich language such as Arabic. Regardless, given the increasing number of Arabic users and the exponential growth of online content, SA in this language has gained the attention of many researches last years, since Arabic raises many challenges because of its derivational, inflectional and agglutinative morphology. The objective of this paper is to promote the performance of Arabic Sentiment Analysis (ASA) by using Deep learning techniques. For that we implement Multi-Layer perceptron model in order to process and classify a dataset (Tweets). In fact, the experimental results prove that MLP as a deep learning model has a better performance for ASA than classical approaches.