In The Beginning, Let There Be The Word Challenges and Insights in Applying Sentiment Analysis to Social Research

Abstract

Sentiment analysis based on lexical corpora is widely employed despite its inherent limitations in capturing nuances such as sarcasm and irony. This research delves into the application of sentiment analysis to political communication. To address the limitations of the Bag of Words methodology, a comparative study of sentiment analysis tools and emotion detection from speech is conducted, using automated speech recognition as a benchmark. Emotion recognition from speech has shown promising results, indicating its potential superiority over other methods [1], [2]. This study uses media material from Polish radio and television broadcasts, focusing on political interviews during a significant period marked by a high-profile assassination attempt. Results indicate challenges at the micro-level, but aggregated data reveals a significant correlation between valence measured from voice and text. While sentiment analysis may lack sensitivity in capturing mourning-related discourse, it proves effective in political communication devoid of such nuances. This suggests that valence in sentiment analysis reflects emotional content derived from intonation fairly accurately.