ISCAP Proceedings - 2023

Albuquerque NM, November 2023

2023 ISCAP Proceedings: Abstract Presentation

Conversational AI: Design AI Dialogue Systems With Object-Oriented Methodology

Thuan Nguyen
The University of North Texas

OpenAI stunned the world when it releases ChatGPT, an artificial intelligence (AI) service, to the public in December 2022. Several months later, Alphabet (Google) introduced its generative AI product, BARD, in May 2023, which was followed by Claude, another large language model (LLM) chatbot owned by Anthropic, a well-known AI start-up. The transformative power of the technology is obvious and overwhelming. It potentially changes various aspects of human life, including work, study, and even daily activities. The advancements of AI, like generative AI, have led to a new technology that can perform multiple tasks done only by human beings before but also take the role of one of the speakers in a conversation. One of the most important applications of conversational AI is creating chatbots that can communicate with human users using natural languages. These chatbots can engage with human users in contextually relevant dialogues. While still an emerging technology, recent advances in natural language processing (NLP) for both goal-oriented and generative conversational AI have resulted in significant progress in developing AI dialogue systems that can serve a broad range of applications, showing an increasingly sophisticated level of natural language understanding (NLU). Worldwide companies of various sizes (big, mid-size, and small) have employed well-known enterprise services like Google Cloud Platform (GCP) Dialogflow CX to build AI dialogue systems used in their daily businesses. Moreover, generative NLP large language models (LLMs) with powerful features that have recently been released, such as OpenAI’s ChatGPT, Alphabet/Google’s BARD, and Antropic’s Claude, can enable users to develop AI chatbots finetuned with specific proprietary data. With these services and advanced generative AI tools, individual users or companies can conveniently develop AI dialogue systems. Now, it is no longer a daunting task to create an advanced chatbot. However, the quality of the conversations that a chatbot or an AI dialogue system can handle becomes the focus of the development process. Therefore, the design of a conversational AI system, including the designs of dialogue flows that the virtual agents of the system can handle, becomes the most important task on which a developer or a team must focus time and effort while developing the system. It is very similar to how a software developer or a team of software engineers focus their time and effort on designing a software application system before writing code to implement it so that the system can run as well expected when it is completed. In this research paper, under the light of the Speech Act Theory proposed by John L. Austin, a well-known Oxford linguistic philosopher, the author proposes a new approach to designing AI dialogue systems applying the Object-Oriented (OO) methodology, a popular paradigm used in developing software application systems. While designing an AI dialogue system using the approach, all significant concepts of OO classes with associated data, i.e., attributes, and behaviors, i.e., methods, will be applied. The conversational intents will be designed with similar structures of class methods. Most importantly, all four OO principles – abstraction, encapsulation, inheritance, and polymorphism will also be employed in designing the AI dialogue flows.