Towards a chatbot for creating trigger-action rules based on ChatGPT and Rasa

In this paper, we propose a novel approach for building a conversational agent for creating trigger-action rules and controlling smart objects inside smart environments, such as a smart home. Our approach integrates ChatGPT, a state-of-the-art pre-trained language model for open-domain dialogue generation, with Rasa, a popular open-source framework for developing task-oriented chatbots. We leverage ChatGPT's abilities to perform Natural Language Processing tasks through prompting and few-shot learning, and Rasa Open Source's features to handle intents, entities, forms, and execute actions. We design Rasa custom actions that invoke ChatGPT's API to process complex customization rules, manage conversational breakdowns and answer questions about the smart environment.