Sentiment Analysis for Arabic Language using Attention-Based Simple Recurrent Unit
Saja Al-Dabet, Sara Tedmori · 2019
With the growing number of people who express their opinions on the web, Sentiment Analysis have become an active research field that aims to analyze and classify the sentiment polarity of opinionated reviews. Recently, Deep Learning models have been extensively used for many Natural Language Processing tasks including Sentiment Analysis. In this paper, the authors propose a Deep Learning model for Arabic language sentence-level Sentiment Analysis. The proposed model represents an integration between an emerged variant of Recurrent Neural Networks known as Simple Recurrent Unit which is characterized by its light recurrent computations, and an attention mechanism that concentrates more on the important parts of an input text. The Simple Recurrent Unit model allows parallel recurrent calculations that lead to enhance the training process in terms of time and accuracy. Experiments were performed to evaluate the performance of the proposed model using the Large Scale Arabic Book Reviews (LABR) dataset. The proposed model obtained state of the art results compared to other Deep Learning models where it achieved 94.53% in terms of accuracy measure with faster execution time.