This AI template will help you summarize a research paper in seconds
As any STEM student would attest, information overload is a major problem in the scientific community. What if you were told that there is a tool to summarize every research article, so you can find the right ones to read? No, we are not talking about the abstract, but something even simpler.
Researchers from Allen Institute for Artificial Intelligence have developed a AI-powered model which sums up the scientific articles in a few sentences. In other words, it condenses a research paper in TLDR (Too Long; Didn’t Read) format so that you can decide which documents are worth reading. It does this by extracting the most important parts of the summary, introduction, and conclusion sections, creating a snippet to describe the document.
How is it possible? Through GPT-3 style neurolinguistic programming (NLP) techniques, which use deep learning to produce human-like text. First, the researchers trained the model on the English language. Then, they created a dataset of over 5,411 abstracts of computer articles and trained the model on over 20,000 other research articles. The result is a nifty tool that can help you sift through hundreds of articles that may seem relevant, to find the ones that suit your purpose.
Summary of the “extreme” research article
The model has been extended to the Semantic Scholar, which is the Allen Institute for Scientific Literature search engine. Now when you search for a research paper here, it will automatically generate a one sentence summary with your results. This summary focuses on the main contributions of the article, removing the methodological details, which are usually summarized in the abstract.
Isabel Cachola, a doctoral student at Johns Hopkins University, believes it will help researchers quickly decide which articles to add to their reading list. âPeople often ask why TLDRs are better than summaries, but the two serve completely different purposes. Since TLDRs are 20 words long instead of 200, they are much faster to browse, âsays Daniel S. Weld, leader of the Semantic Scholar research group at the Allen Institute for AI and professor of computer science at the University of Washington.
TLDR functionality is now available in beta for nearly 10 million articles in Semantic Scholar. It’s limited to the IT realm for now, but you can expect it to move into other areas soon. Try it yourself here, then spread the word to your science friends!