InterSearch A proof of concept application for structured query building with LLM capabilities

Traditionally, researchers seeking academic papers were required to spend extensive hours searching through university libraries to locate relevant materials. With the advent of the Internet, digital libraries and academic organizations developed online repositories equipped with their own search engines, enabling users to retrieve research papers by formulating keyword-based queries. However, constructing precise search queries remains a challenge, as each database follows a unique syntax and glossary. In the era of Artificial Intelligence (AI) and Large Language Models (LLMs), research discovery has been further revolutionized. Modern LLM-powered applications, such as ChatGPT and Gemini, can assist researchers by generating well-structured search queries tailored to specific academic databases. For instance, databases like PubMed provide detailed search syntax guidelines, and an LLM can leverage these rules to construct optimized queries. In this thesis, we present an application that harnesses LLMs to streamline the research process by automating query generation. Our system allows users to input their research intent in natural language, after which the LLM formulates a refined search query aligned with the specific database’s search framework. Additionally, the system provides an explanation of the query structure, enhancing user comprehension of how the search parameters were determined. This approach significantly reduces the time and effort required for academic research while improving search precision and efficiency.