Role-Play Prediction using Ontology-Based Graph Convolutional Network Model

Current applications of large language models often assign tasks without consideration of how LLMs understand a given prompt.Simple commands sometimes do not guarantee desired responses, as LLMs are systems based on mathematical modeling and cannot cognitively be capable of understanding commands.Hence, a method is required to guide LLMs in performing tasks appropriately.This paper presents a method to develop model-based automation of role selection supported by ontology.This can allow for more accurate and relevant role recommendations than if done manually.As such, this optimization at hand improves the performance of LLMs for specific tasks and overcomes the limitations of previous studies that define the roles by hand.