AI-BASED COMPUTATIONAL PSYCHOLINGUISTICS FOR EMOTION DETECTION: A LITERARY STUDY OF CHARACTER PSYCHOLOGY IN ENGLISH FICTION

Abstract

Computational psycholinguistics has enabled deeper and more systematic interpretation of linguistic markers that encode emotions, cognition, and psychological states. This study investigates how AI-based emotion-detection models can be applied to English fiction in order to examine the psychological depth of characters through their utterances, narrative descriptions, and dialogue patterns. Using transformer-based architectures such as BERT, RoBERTa, and DistilBERT, the study develops a psycholinguistic framework that identifies emotional cues through lexical affect, syntactic patterns, and contextual embeddings. A curated corpus of 12 English novels from the nineteenth and twentieth centuries was annotated using well-established emotion taxonomies, including Plutchik’s eight-emotion model and Ekman’s six-emotion framework. The analysis quantifies emotional volatility, character affect trajectories, and narrative emotional weighting. The proposed framework demonstrates that modern computational models can uncover implicit psychological states that are often inaccessible to traditional literary criticism. Results show an average improvement of 14–18 percent in emotion-classification accuracy when contextual embeddings are combined with psycholinguistic features such as sentiment polarity, concreteness scores, and cognitive-process markers. The findings highlight the promise of AI-driven literary analytics in advancing digital humanities, enabling objective psychological profiling of characters, and offering new pathways for literary interpretation grounded in quantifiable linguistic evidence.