AI system analyzes literature with symbolic rules
Summary and headline written by AI from the source article. How we work
Yuhang Du of Boda College of Jilin Normal University developed SymbolFusion, an artificial intelligence framework that analyzes literature by identifying symbols and tracking emotional shifts.
The system focuses on recurring imagery, symbolic relationships, and emotional changes within texts, addressing limitations of existing AI that often only assesses sentence sentiment. SymbolFusion builds a graph connecting imagery, concepts, and emotions, using specific rules to ensure valid connections and track how emotions evolve throughout a passage.
The framework uses a “literary compatibility tensor” to filter relationships, accepting only those that meet pre-defined criteria based on literary conventions. This ensures the AI focuses on meaningful connections rather than random associations. Du evaluated SymbolFusion using a corpus of 3,120 English literary passages, finding it achieved 91.3 percent agreement with human annotations of symbolic graphs and a high level of consistency in emotion tracking.
The research demonstrated that incorporating explicit literary rules improves the AI’s performance compared to general-purpose language models. While SymbolFusion currently works best with passages up to 1,000 words due to memory limitations, Du positions it as a tool to assist, not replace, human literary analysis, offering testable hypotheses about a text's structure and emotional impact.
