Specialized AI Chatbot for Science and Technology Information

Overview

The project recognizes a critical need for expanded training data due to the current shortage of AI models tailored for science and technology. By promoting intelligent and efficient task operations for all relevant personnel, the service seeks to improve the practical usability of the organization's vast data resources. Consequently, the project is dedicated to constructing a domain-specific ultra-large language model and deploying diverse generative AI solutions customized to meet professional requirements.

Project description

The project leveraged extensive datasets from the 5th Science and Technology Basic Plan managed by the Korea Institute of Science and Technology Planning and Evaluation—comprising 10,000 meeting agendas, 5,000 trend documents, and 20,000 internal reports. By developing an in-house conversational AI and a fine-tuned pilot model, the service delivers tailored science and technology insights to a wide range of stakeholders. Specifically, it assists businesses with optimal partner matching, supports researchers in navigating complex regulations and drafting proposals, enables government ministries to execute data-driven policy planning, and provides the general public with science-based solutions for social issues. Furthermore, the establishment of a document-search and Q&A-responsive GPT allows users to access precise information with significantly enhanced speed and accuracy.

Key technological features

• Data Preprocessing: Preprocessing approximately 1,839 passages of internally held data
• Training and Indexing: Training and indexing research report data into SearchGPT

Expected outcomes

By developing a suite of user-customized generative AI services, the project establishes a robust utilization framework for the proprietary data held by the Korea Institute of Science and Technology Planning and Evaluation (KISTEP). The service is expected to maximize the intelligence and operational efficiency of tasks performed by diverse stakeholders across the science and technology sector.

Launch year: 2023

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=29185