Reproduce “MBTI Personality Prediction for Fictional Characters Using Movie Scripts” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yisi Sang, Xiangyang Mou, Mo Yu, Dakuo Wang, Jing Li, Jeffrey Stanton
Researchers developed a new benchmark dataset, Story2Personality, and a Multi-View Multi-Row BERT model to automatically predict MBTI personality types of fictional characters from movie scripts. The model demonstrated a 2-3% F1 score improvement over baselines by effectively integrating both character dialogue and scene descriptions.
Key insights
- Integrating both dialogue and scene descriptions (multi-view information) consistently improves personality prediction model performance compared to using either source alone.
- The personality prediction task for fictional characters from movie scripts remains challenging for AI models, with human performance being substantially higher (>95% F1 score) compared to model performance.
- Long-text processing and capturing the full context of character development over an entire movie script are critical challenges for current NLP models.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.