Conference paper Open Access
In this article, due to the importance of sentiment analysis, it has been strived to design a framework that is capable to distinguish the polarity of the opinions in the Persian language. For reaching out the goals of the article, a Persian corpus is used and three baseline models such as Support vector machine, Naive Bayes and Stochastic gradient descend have been selected in order to classify the sentence-level of a dataset. Then common deep learning models that are suitable for text mining like long short-term memory and convolutional neural network applied on embedded vectors. Through this paper, some methods based on data augmentation, word embedding, and word representation have been used also. Consequently, it will be shown that the proposed deep learning models work with high accuracy in comparison with the baseline algorithms.