Semantic Analysis of Social Aspects in Textual Data: Exploring Agency and Emotions with NLP Techniques

With the rise of digital communication, understanding the social aspects embedded in textual data has become increasingly important. This thesis focuses on the semantic analysis of social dynamics, specifically targeting negative emotions and agency, using advanced Natural Language Processing (NLP) techniques, including BERT-based models. We also utilize tools like SpaCy and LIWC to analyze dependency structures and detect emotional content within the text. Our study applies PageRank-like marker projection (PLMP) techniques to explore word-level contributions to emotional perception and agency. Additionally, we conduct a correlation analysis to investigate the relationships between linguistic features, social roles, and emotional markers, uncovering important patterns in the expression of power dynamics. Through these methods, we provide a framework for better understanding how language shapes social interactions. This research underscores the importance of detailed word-level analysis in comprehending the complexities of social interactions within text.