Enhancing Chatbot Interactions Through Personality-Driven Style Transfer
Résumé
In recent decades, the conversational agent's (CA) development has sparked a lot of attention. Thanks to recent deep learning and Natural Language Processing (NLP) techniques advancements, AI powered chatbots can now interact with humans, understand their needs, and reply to them in a natural way. However, the emphasis on CAs’ designing and development has shifted from only considering their technical capabilities. There is now a greater focus on enhancing their social and conversational skills, coupled with the integration of distinctive personalities. This trend in AI technologies toward the social is motivated by the idea that humans respond to machines in the same way they would to a human. In the field of psychology, personality is described as the mixture of a person's behavior, motives, and traits. Therefore, communicating with others who have similar personality types is straightforward since they use the same communication channels. Drawing inspiration from this understanding, we introduce an approach for endowing personality into a language model-based conversational agent. The distinctive feature of our system lies in its real-time personalization, addressing the complexities of human communication influenced by emotions and personality. The system adeptly adapts interactions based on the user's present emotional state and inherent personality traits. The system achieves this high level of personalization through the seamless integration of three key modules: Emotion and Personality detection modules, and a Text Style Transfer module. Together, these modules empower the conversational agent to tailor its responses in real-time, creating a more engaging and personalized user experience.