Harnessing NLP and AI to decode political discourse: Speech patterns, sentiment analysis, and public perception

Abstract

Using natural language processing (NLP) and artificial intelligence (AI), this study analyzes the frequencies of words and phrases in political leaders’ speeches to track patterns in political discourse. The objective is to identify language patterns, sentiments, and topics of political addresses using state of-the-art methods like automatic transcription (Whisper), Bidirectional gated recurrent unit (GRU) for sentiment analysis, and BERTopic. Through the use of Whisper’s state-of-the-art transcription service, we were able to transcribe the political speeches into machine-readable text, which in turn provides for other types of analysis. Bidirectional GRU classifies sentiment as positive, negative, or neutral with the aim to study how politicians use sentiment to manipulate their listeners. Furthermore, we use BERTopic for tracking the evolution of rhetoric, key trend summarisation, and topic mining and analysis. It illustrates how politicians employ discursive strategies and epilinguistic elements to manage the public mind and reality. Achievements and objectives are framed with positive and defensive emotions aimed at threats or criticisms. The emotional grab of it all is still important. It locates in these the thematic coherence and shifting sentiment that lie at the heart of political storytelling. It shows how political communication is evolving to stay relevant in the digital media age and delivers language – even real-time language pattern tracking – via the use of AI and big data. Further study is needed of multimodal and flexible techniques for analysing political discourse across languages and time periods.