Analysis of Fictional Character Backstories Using Natural Language Processing and Deep Learning

With the overwhelming successes of industries like Marvel and DC over the years, fictional characters have become more and more popular among the general mass. However, generation of new fictional characters do not seem to be as appealing as before. This paper revolves around the idea of generating fictional character backstories using GPT-3.5. A dataset consisting of AI generated pre-existing and newly created fictional characters has been prepared. A sequential neural network model was designed by the authors and pre-trained using the popular Superheroes NLP dataset from Kaggle. The proposed model and some state-of-the-art pre-trained text classification BERT models have been compared side by side. The AI generated dataset has been used as the test data to see how the models perform in identifying the moral alignment of the characters. A comparative analysis is also done between the two datasets.