Random forest and recurrent neural network for sentiment analysis on texts in Bulgarian language

The purpose of this paper is to further explore and search for the most accurate methods and algorithms to use in sentiment analysis on comments in Bulgarian language. In the previous works of the authors were applied standard algorithms like Naïve Bayes, Support Vector Machines and Logistic Regression on two self-developed databases with customers’ reviews in Bulgarian. The objective of the current project is to find weather Random Forest and Recurrent Neural Network could give better accuracy in sentiment prediction than the other algorithms or reduce the difference in the results of the two databases. The aim of the authors is to find the most accurate model for sentiment analysis in Bulgarian language, given the specifics of the language.