Artificial Intelligence Can Help Readers Digest Scholarly Papers More Efficiently

The volume of research papers in the literature and the lack of time to read them can be a challenge for scientists. A study from Learned Publishing reports that it takes an average researcher from the United States 32 minutes to read an English-language paper. Another study from Publications reports that if English isn’t a researcher’s first language, it takes nearly one hour or longer to read an article.
Paper Digest, founded by postdoctoral researchers Yasutomo Takano and Christian Mejia, is a new resource that skims and summarizes manuscripts for readers. This artificial intelligence tool is designed to pull key central concepts from full-text articles and lists them in a single-page summary.
Paper Digest has a simple web interface, so users do not need to download or install anything. After entering an open-access article’s DOI or the URL for the full-text PDF, the tool returns a page summary in approximately 10 seconds.
Takano admits that currently, Paper Digest’s algorithm is best at analyzing manuscripts with clear headers, such as those used in original research articles published in Science Translational Medicine (STM) journals. Review articles, editorials, and articles from disciplines that do not follow the standard STM article structure are more difficult to analyze and the summaries are not as accurate. Takano also explains that the algorithm currently works best with text only; if key concepts of papers include math equations, the algorithm may fail to analyze them.
The Paper Digest team says that their focus is on improving the algorithm. Takano says for that to happen, lots of datasets are needed to feed the machine learning. A new feature was recently added to gather feedback that allows users to press a “like” button next to a helpful extracted sentence. The developers also encourage users to communicate when articles in particular disciplines do not provide accurate summaries.
We at the Ochsner Journal understand that as technology advances, more artificial intelligence-based tools will be available. We will continue to monitor the developments at Paper Digest and other tools that may potentially assist with scholarly research.