A deep learning approach to classify fictional and comic characters into four alignment categories (Good, Bad, Neutral and Antihero) based on their backstories using natural language processing (NLP) has been presented. Our methodology incorporates comprehensive text preprocessing, including data cleaning, WordNet-based augmentation and dataset balancing techniques. The provided classification model employs a neural network architecture with vectorizers, embedding layers, global max pooling, and regularized fully connected layers, optimized using Adam with early stopping along with it to enhance the user experience, a cosine similarity retrieval system is integrated which identifies the most similar character in our database based on input descriptions and provides that along with the alignment. The model achieves approximately 91% validation accuracy which is better compared to all previous state of art models, demonstrating its effectiveness for text-based classification tasks. Beyond entertainment applications, this approach shows potential for adaptation to criminal profiling and behavioral analysis which can be helpful for analyzing class of a criminal profile using a similar NLP model, where textual descriptions of backgrounds and psychological factors could predict behavioral patterns