Optimizing fuzzy logic based text sentiment analysis through machine learning

Technologies advancement introduced new challenges in the field of human communica-tion analysis. Sentiment analyses has found its usage in various fields, such as marketingand psychology. Traditional text analysis systems rely on neural networks. Nevertheless,learning algorithms might be quite time consuming. Fuzzy logic is a light weighted mecha-nism that interacts with abstract and fuzzy concepts, such as human thinking. This articleresearches the way how fuzzy logic may be used along with machine learning algorithms tocreate a better system solution for text sentiment analysis.To simplify fuzzy inference system modeling for this article, a prototype knowledge basewith emotionally charged words ranging from negative to positive sentiment was devel-oped. The fuzzy system evaluates sentiment using three linguistic variables – “negative”,“neutral”, and “positive” – with Gaussian membership functions for negative and positivevalues and a triangular function for neutrality, that allows to improve nuanced sentimentclassification.The fuzzy model utilizes the “min” aggregation method along with center of gravitymethod to calculate text sentiment score based on the input text and available knowledgebase. As the full model is hybrid it combines fuzzy inference system with machine learningalgorithm to optimize membership functions. The algorithm updates membership func-tions parameters such as height and width, that results in more adaptable and accuratefuzzy inference system.The future work is going to be focused on the development of more complicated hybridsystem. This also includes researching different machine learning algorithms and optimiz-ing more parameters to make the system more robust.