Akademik Veri Yönetim Sistemi
KNOWLEDGE AND INFORMATION SYSTEMS, cilt.67, sa.9, ss.7319-7354, 2025 (SCI-Expanded, Scopus)
Chatbots have become increasingly popular by transforming interactions across numerous fields. As the technology behind chatbots has rapidly developed, new methodologies have arisen, each contributing unique strengths and addressing different challenges. This paper systematically reviews the methods used in chatbot development from 2019 to 2024, comprehensively analyzing the studies. We categorize the techniques into three main groups: machine learning (ML)-based, deep learning (DL)-based, and large language model (LLM)-based methods. We present a broad and inclusive survey by exploring the foundational principles of chatbot technologies and their applications across diverse domains such as education, healthcare, and interviews. Our analysis reveals that while traditional ML-based methods remain widely used, DL models are gaining prominence for handling complex tasks, and LLM-based systems are advancing the field by offering more coherent, contextually aware responses. However, challenges remain, especially in ethical concerns like hallucination and privacy-preserving technologies, particularly with LLMs. The paper also identifies gaps in existing research, notably the need for improved privacy-preserving mechanisms and better strategies for mitigating hallucinations in chatbot responses. Future research directions are suggested to address these challenges, particularly in developing LLM-based chatbots, with a focus on enhancing privacy, accuracy, and ethical standards.