Sentiment Analysis Approach Based on Combination of Word Embedding Techniques

Sentiment analysis is a field of research that attracts the attention of companies and governments to understand the opinion of the client and citizens, about products, services, policies and more other. With the increased volume of user-generated content on the Web, especially social networks, textual information becomes freely accessible and with a gigantic quantity, which requires powerful automated analysis tools to extract such kind of information (positive or negative sentiment). In this paper, we present sentiment analysis approach depends on pre-trained word embeddings, a frilly high-quality word representation vectors, namely, AraVec and fastText models, and we proposed a combination of the two models, based on vectors concatenation of both models. Sentiment classification was executed employing six different machine learning algorithms, we find that in most of the cases, our proposed method achieves the best results in terms of accuracy, especially with NuSVC classifier which is a type of SVM.