Is this research paper too long to read? AI can sum it up for you

Most students would attest to spending countless hours reading research papers. Researchers at the Allen Institute for Artificial Intelligence have developed a new AI-based model that summarizes key arguments from scientific papers. So you don’t have to drown in the text.

This free tool, known as TLDR (the Internet acronym for “Too long, did’t read”), condenses a study into a concise, one-sentence summary. It does this by focusing on the salient information of the article in the summary, introduction, and conclusion sections. This software removes things like methodological details, which are usually summarized in the abstract.

Researchers and engineers at the nonprofit Allen Institute for Artificial Intelligence (AI2) in Seattle, Washington, used GPT-3-style neurolinguistic programming techniques to create this AI-powered model. They trained it with a dataset of over 5,411 computer science papers with corresponding abstracts, some written by the team themselves and others by a class of undergraduate students from the University of Washington. They further improved the performance of the TLDR model by collecting training examples from 16 other areas.

The tool has been enabled for search results on Semantic Scholar, a search engine created by AI2. At the moment, it is only available in beta version for more than 45 million articles in English in the fields of computer science, biology and medicine. The researchers are currently improving the software so that it will expand to more languages ​​and domains in the coming months.

For Daniel S. Weld, who manages the Semantic Scholar group at AI2, one-sentence summaries of research papers could help scientists make quick and informed decisions about which papers are worth reading in more detail. “People often ask why TLDRs are better than digests, but the two serve completely different purposes. Since TLDRs are 20 words instead of 200, they are much faster to scan,” he said in a statement.

It’s short and sweet… and funny

AI2’s TLDR software isn’t the only scientific synthesis tool that cuts through academic jargon. Academics recently shared AI-powered summaries of their research papers on Twitter that “a sophomore can understand,” courtesy of tl;dr papers. This website was created by software engineers Yash Dani and Cindy Wu, according to Professor Michelle Ryan, director of the Global Institute for Women’s Leadership at Australian National University.

Ryan posted on his Twitter account the AI ​​summary of one of his articles on the “cliff of glass”, a form of gender discrimination in which women are appointed to leadership positions when companies are most exposed. at risk of failure. But that’s not how the AI ​​would explain it. According to tl;dr papers, “The Glass Cliff is a place where a lot of women get put. It’s a bad place to be.

Dr. Laura Sockol, clinical psychologist and associate professor at Davidson College, also shared the summarized version of her article, “Improving Quantitative Skills and Attitudes in Clinical Psychology Courses.” For the machine, this research paper focuses on the fact that “when students take psychology courses, many don’t like to learn statistics.” Why is that? Because “they are bad at it and they are afraid of it”.

Although the summaries provided by tl;dr papers received an enthusiastic response on Twitter, the website has been labeled “under maintenance” ever since. It looks like students will now have to skim through the most important parts of a research paper the old-fashioned way – by reading them. – AFP Relax News


Source link

Comments are closed.