AI-powered Chatbot for Information Service at Klabat University by Integrating OpenAI GPT-3 with Intent Recognition and Semantic Search
Semmy Wellem Taju, Green Ferry Mandias, Jimmy Moedjahedy, Andria Kusuma Wahyudi, Reymon Rotikan, Edson Yahuda Putra · 2023
The rapid advancement of Artificial Intelligence (AI) in recent years has transformed numerous sectors, including information services. This study introduces Unklabot 1.0, an innovative AI-powered chatbot specially designed and proposed for information services at Klabat university. Unklabot 1.0 was developed to take advantage of an advanced large language model (LLM), namely OpenAI GPT-3 (text-davinci-003), and integrate it with intent recognition and knowledge-based context understanding using semantic search technique to provide accurate and efficient question answering capabilities about Klabat University. To enhance Unklabot's performance, intent recognition is integrated into the Unklabot system. Moreover, this intent recognition helps identify the purpose behind user queries, allowing the chatbot to understand user intentions more accurately. Furthermore, semantic search technique is employed to enhance the capabilities of the Unklabot by understanding the meaning behind user queries and retrieve information in external knowledge based on the closest context or chunk. Utilizing OpenAI GPT-3 API, Unklabot can processes natural language inputs, providing comprehensive and contextually relevant answers to user inquiries. This research successfully implemented the OpenAI GPT-3 API in information services, offering a valuable tool to increase the engagement of AI usage in educational institutions such as Klabat University.