Authors: Muhammed Yusuf Sermet, Ibrahim Demir – University of Iowa
Title: Development of an Intelligent Hydroinformatics Chatbot and Virtual Educational Assistant for Enhancing Learning and Research in the Hydrological Domain
Abstract: The hydrological domain requires access to large volumes of data for research and professional use cases. To facilitate this, the development of an Intelligent Hydroinformatics Chatbot and Virtual Educational Assistant is presented, which enable users to interact with hydrological APIs and educational resources in natural language. To power both components, a large language model (i.e., GPT-3) is adopted and fine-tuned with the developed knowledge graph and a curated training dataset. The Intelligent Hydroinformatics Assistant can parse APIs that follow OpenAPI specifications, fine-tuning the model for available endpoints, parameters, data types, and other relevant metadata, while manual processing is required for others to reflect API details on the provided template. The Virtual Educational Assistant is created using the same framework and can allow instructors to generate virtual assistants based on all available content on the learning management system, such as Canvas, providing students with conversational study tools that promote self-learning. The project focuses on two main components: (1) Intelligent Hydroinformatics Assistant and (2) Virtual Educational Assistant. By adopting and customizing a large language model, such as the one powering prominent conversational assistants (e.g., ChatGPT), the developed framework will improve the accessibility and usability of hydrological data for research and professional use. The Virtual Educational Assistant will provide students with a conversational study tool to enhance learning in the hydrological domain.