MBTI Personality Prediction for Fictional Characters Using Movie Scripts
This addresses the problem of narrative comprehension for AI researchers, but it is incremental as it builds on existing personality prediction tasks with a new dataset.
The paper tackled predicting MBTI or Big 5 personality types for movie characters from narratives, finding that existing models perform poorly, and proposed a multi-view model that improves results by incorporating verbal and non-verbal descriptions.
An NLP model that understands stories should be able to understand the characters in them. To support the development of neural models for this purpose, we construct a benchmark, Story2Personality. The task is to predict a movie character's MBTI or Big 5 personality types based on the narratives of the character. Experiments show that our task is challenging for the existing text classification models, as none is able to largely outperform random guesses. We further proposed a multi-view model for personality prediction using both verbal and non-verbal descriptions, which gives improvement compared to using only verbal descriptions. The uniqueness and challenges in our dataset call for the development of narrative comprehension techniques from the perspective of understanding characters.