Development of an AI-Driven System for Scientific Article Context Extraction and Knowledge Graph Visualization

Literature review plays a crucial role in exploring prior research, serving as a foundation for subsequent studies. It outlines the content of scientific articles, including background, objectives, and methodologies. This research aims to develop a system for extracting scientific articles and visualizing knowledge graph based on Large Language Models (LLMs) to facilitate the literature review process. Knowledge graph, represented as an information network of nodes and edges, enables data visualization and maps relationships between scientific documents, offering new perspectives for writing. Previous research has demonstrated the potential of knowledge graph in fields like education and cybersecurity. In this study, knowledge graph is developed using a waterfall model. The integration of LLMs, such as ChatGPT, accelerates information extraction by providing keyword recommendations and document summaries. System evaluation is conducted through user questionnaires to assess the effectiveness of knowledge graph visualization in supporting academic writing. The research identifies six functional requirements and one non-functional requirement for the system. System testing includes black box testing to evaluate functionality and compatibility testing using SortSite. A questionnaire with eight respondents shows three strongly agree that the system is helpful and beneficial. This study is expected to significantly contribute to AI-based technology development to support academic activities.