Categorical text generation is an upcoming field within natural language processing focusing on producing text tailored to specific categories or domains. The goal of categorical text production is to produce text that complies with the requirements and characteristics of the chosen category while also adhering to the grammatical and structural conventions of the language. Text classification works across multiple data categories, and the synthetic data generated usually adheres to a preset pattern and ignores the text’s nuanced emotional content. The proposed approach begins with using BERT to get the text’s deep contextual embeddings. Using these text embeddings, the underlying semantics of the target category can be better understood by adding a task-specific layer that places the sentences in the closest corresponding class. After that, the architecture and settings of the GAN are optimized through evolutionary computing techniques to guarantee that the generated text conforms to the constraints and characteristics specific to each category.