Read Literature Like a Map

Reviews and bibliometrics can assist researchers in acquiring structural knowledge. However, reviews may not always be available for every specific topic, and bibliometrics fails to consider the actual content of papers. AI tools can facilitate the acquisition of detailed information more easily, but their reliability can be questionable, and researchers must first identify the relevant papers. This study introduces a framework, literature map, that integrates the actual content of papers into vectors and clusters them. These clusters are then translated into a map view, with customized summarizations generated by a large language model to reveal structural knowledge. Graph knowledge and text chunks are extracted from the internal text of the literature and integrated into a graph database, enabling both global and local searches for detailed knowledge. The literature map framework is both universal and customizable, and can serve as a complement to reviews and bibliometrics. A video example has been uploaded to YouTube at www.youtube.com/watch?v=phkr9Efv9fI.