Hyper-scale AI-Driven Integrated Research Support Services

Overview

The need to improve research quality has emerged by creating a Generative AI-based environment that minimizes the significant human and time resources typically spent on literature review and data analysis. An integrated research support service based on a Hyper-scale AI platform will provide the foundation for an efficient research environment. This is expected to contribute to national development and social value by enabling timely responses to major national tasks and social issues.

Project description

Researchers can engage in dialogues with the AI regarding new research topics and methodologies. They can utilize auxiliary services such as document review, translation, and the generation of meeting minutes or presentations throughout the lifecycle of research planning, execution, and performance dissemination.

• Proposal: Suggests new themes through research exploration and topic validation.
• Efficiency: Enhances focus and quality via report generation and brief summarization, enabling better data analysis and resource forecasting.
• Collection: Gathers diverse materials through similarity analysis with related studies, assisting in the sharing of research outcomes.

The service aims to innovate the research environment by reducing the time researchers spend on policy studies, which currently averages over a year. Regarding Research Planning, the platform supports decision-making by suggesting research topics and methodologies and generating preliminary research proposals. For Research Execution, it provides services such as research proposal analysis, report summarization, and similarity mapping with related studies. Additionally, researchers can access Auxiliary Support features like automated meeting minutes and grammar/context error detection. Finally, to facilitate the Dissemination of results, the system offers verification of research regulations and the automated generation of presentation materials.

The project involved developing a conversational service interface and establishing cloud infrastructure. A Data Lake was constructed and used for training/retrieval, incorporating data from the National Knowledge Information System (NKIS), the National Law Information Center, internal regulations of 11 affiliated research institutes, and news articles.

• Data Collection & Pre-processing: Original data—including research reports, policy materials, periodicals, and seminar papers—is collected and processed. Through a "Data Pre-processing Pipeline," the team completed RAG design/optimization testing, data vectorization, and the development of systems for data collection and real-time updates.

Key technological features

• Hyper-CLOVA X-based Platform: Utilization of CLOVA Studio to develop specialized LLM (Large Language Model) services for research assistance
• Data Infrastructure: Construction of a dedicated Data Lake for research data and implementation of a RAG (Retrieval-Augmented Generation) based search system

Expected outcomes

The provision of an integrated research support service using Hyper-scale AI significantly improves research efficiency. By automating data analysis and processing, researchers can save time and focus more on core analysis and strategic thinking. Building an environment where vast amounts of research data can be learned through AI improves the overall quality of research results and contributes to better policy decision-making. Furthermore, the digital platform makes collaboration between researchers from different fields easier, and convergence research becomes possible through AI trained on multifaceted data.

In addition, resources such as manpower and time can be utilized more effectively. Hyper-scale AI allows for the preemptive filtering of similar or redundant research, thereby shortening the overall research period. Through a standardized learning data and service structure, the resources required for repetitive data training and service construction are minimized, ensuring a sustainable and cost-effective research ecosystem.

Launch year: 2024

Attribution

Case study courtesy of the National Information Society Agency, Republic of Korea, Apolitical's Content Partner. Sourced from Use Cases of Public AI Service, Vol.2, https://eng.nia.or.kr/site/nia_eng/ex/bbs/View.do?cbIdx=31975&bcIdx=29182&parentSeq=29219