In order to adequately analyze the emotional and rhetorical content of political speech, which is becoming increasingly important in the process of forming public opinion, sophisticated methods are required. By utilizing artificial intelligence (AI) and natural language processing (NLP), this research investigates the method of analyzing the sentiments expressed by political leaders through their speeches. We construct and analyze classical machine learning models, deep learning approaches, and transformer-based architectures in order to categorize the attitudes that are embedding themselves in political statements. We have assembled, preprocessed, and analyzed a varied dataset consisting of speeches delivered by political personalities from throughout the world. The findings indicate that transformer models perform better than other models when it comes to collecting contextual sentiment, which leads to the provision of useful insights into political communication. The research makes a contribution to the field of political natural language processing by comparing the efficacy of several AI techniques.

Keywords: NLP, Deep Learning, Machine Learning, BERT, CNN, LSTM, Random Forests, SVM