This project investigates the application of NLP and machine learning to produce computational estimates of the Big Five personality traits and emotions contained within the dialogue and behavioural descriptions of fictional characters in film and television screenplays. A system was developed as an exploratory feasibility study to assist individual screenwriters in evaluating how characters resonate with audiences through their narrative portrayal. By integrating NLP models and sentiment analysis tools into a system pipeline, the study was able to help establish a greater insight into the boundaries of what pre-trained models can achieve when applied to a fictional context, an area that is predominantly constrained by domain mismatch. This represented a fundamental barrier that could not be resolved within the scope of this project.