AutoSurveyGPT: GPT-Enhanced Automated Literature Discovery
Chang Xiao · 2023
In this work, we introduce AutoSurveyGPT, a novel framework for literature discovery. Designed to accommodate brief user-provided descriptions of academic papers, ideas, or proposals, this system is capable of autonomously extracting keywords for subsequent exploration within scholarly search engines. By leveraging large language model like GPT-4, the system further evaluates the relevance of the retrieved papers to the user-provided idea. This process, based on examining the introduction and related work sections, drives a repeating cycle of creating new keywords and finding more papers. The system generates a list of related papers, effectively aiding researchers in their search for relevant work. The open-source code for this tool is available on GitHub https://github.com/a554b554/AutoSurveyGPT.