Emotion-Aware Movie Genre Classification Using Dialogues

Abstract: This paper presents an emotion-aware approach to movie genre classification that leverages the linguistic and affective patterns embedded within dialogues. Unlike conventional genre prediction models that depend on metadata or plots, this study utilizes a fusion of lexical and sentiment-based features to predict movie genres. The system combines TF–IDF representations with emotion cues derived from VADER sentiment analysis, thereby enhancing contextual and affective understanding. Multiple machine learning models, including Naive Bayes, Logistic Regression, Linear Support Vector Machine (SVM), and an Ensemble classifier, were trained and compared. The best-performing model, an Ensemble combining Logistic Regression and SVM, achieved an overall accuracy of 53.23% across ten genres. The findings demonstrate that emotion-informed textual features significantly enhance the accuracy and interpretability of movie genre classification systems.

Keywords: Movie dialogues · Emotion analysis · Genre classification · Sentiment features · Machine learning · Natural Language Processing (NLP)