Motivation: Artificial Intelligence systems excel at scanning through, organizing, and summarizing large natural language databases. The NIH estimates the PubMed database alone contains more than 36 million scientific articles, with another million entering the database per year - meaning that every single minute, the database expands by two articles. These numbers are beyond the capacity of human reviewers to handle independently, which is why we’ve built an AI-assist for literature review tasks.

Introducing, SearchBio! SearchBio is capable of ingesting scientific queries, scanning tens of millions of articles for relevant literature citations, then summarizing articles of interest for rapid understanding of an article’s methods, thesis, and supporting data.

Read on below for how to get started.

1. Begin by navigating to the search bar and entering a query of interest: e.g. “What is the impact of tryptophan on sleep?”

Users can filter by field of study, publication type, publication year, and whether or not the search result returns an open access PDF.

1. After searching 200 million papers, SearchBio returns the 20 most relevant publications matching your query, with their abstracts summarized in a single sentence. This allows for quick scanning for relevant papers of interest.

1. Additionally, users may also prompt SearchBio to generate a bullet point overview of the Main Findings, Limitations, or Methodology of the paper. In order to generate this info, remember to check the left-hand box next to the paper title!

1. See a paper that looks interesting? To dive deeper, navigate back to the left-hand check box, and click the speech bubble to ‘Ask More’.

This re-directs the user to DocChat, automatically ingesting the chosen PDF into our interactive Q&A app. Note: the paper MUST have an associated PDF for this to work!