A Machine Learning Guided Path for Optimal Literature Review

This paper introduces a novel machine learning framework to address the challenge of optimizing literature research by identifying the optimal path. To create dataset and ensure the versatility of the solution for different applications, we developed an online scraping tool designed to extract articles from ResearchGate based on a specific search query. The proposed machine learning model leverages contextual embeddings and graph theory, translating intricate scholarly work into informative steps for one to go wider rather than deeper in their research. By employing a Christofides approximation of the Traveling Salesman Problem algorithm, our model efficiently navigates through more than 1000 article embeddings. We prove that the resulting path not only accelerates the knowledge gaining process, but also evidently diversifies the findings. Moreover, we evaluated multiple PDF reader libraries to arrive at the most suitable one for the purpose. This adaptability allows the framework to be applied not only to scraped articles, but also to those stored as PDF files, giving an option for multiple data sources. In conclusion, this paper presents a transformative approach for literature research optimization, equipping researchers with a potent tool to efficiently explore articles.