Each day, hundreds of papers are uploaded to Cornell University’s Arxiv.org. Yet, technical jargons in research papers are scaring away audiences with zero backgrounds. Furthermore, the use of technical jargons in research papers has been continuously growing(https://elifesciences.org/articles/27725). Additionally, due to the fact that average audiences don't understand how these scholarly papers relate to them in real life, they’d have no motivation to read them from the first place, even if the contents were simplified.
- 16.2 to 18M visitors during the past 3 months visited Arxiv.
- Yet, 47% of visitors navigated away from the site after viewing only one page.
We can utilize GPT-3 to break the ice between average audiences and technical papers. This can be accomplished by generating a creative real-life applications of the proposed method from research papers, as well as summarization that a 2nd grader can understand. To raise public awareness, we want to create a Twitterbot and newsletter subscription that shares weekly summarized papers, in conjunction with a recommendation system that allows users to stick with their subject of interest.
We created a website at bytesizearxiv.org which continuously takes in papers from Arxiv and summarizes them into bullet points, and puts these summarizations into a PostgreSQL database. The website then reads off the database and contains the summarizations of thousands of papers from Arxiv.