Sentiment and complexity analysis on two databases in Bulgarian language – final estimation

Abstract

The objective of this paper is to explore in depth and find the reason that lies behind the difference in the results that two databases with user reviews in Bulgarian language give after the application of algorithms like Logistic Regression, Support Vector Machines and Naïve Bayes, as well as Random Forest and Recurrent Neural Network. In the author's previous works were created two databases with customers' comments in Bulgarian language and after that were applied the stated algorithms. The purpose of the current project is to find the reason for the different final accuracy results the two databases give, by making a complexity analysis of the language used in them. In addition, the authors aim to find the most accurate model for complexity analysis in Bulgarian language, given the specifics of the language.