How NeuroBriefer works

The problem NeuroBriefer is trying to solve

Most researchers do not struggle because they cannot find papers. They struggle because every week there are more papers than they realistically have time to read. NeuroBriefer is built for that triage problem: deciding what deserves attention first.

Current literature focus

NeuroBriefer currently focuses on peer-reviewed neuroscience literature. The app surfaces PubMed links and PubMed-style article metadata where available, such as PMID, title, authors, journal, publication date, abstract or summary, and article type.

Preprints are not currently included. I would like to add optional preprint support in the future, but it should be a choice because some researchers want early preprint awareness while others prefer weekly briefings limited to peer-reviewed papers.

Keywords and conceptual matching

Users provide keywords and research interests. NeuroBriefer uses those terms to represent the conceptual area a researcher cares about. In the current app, profile keywords are converted into an embedding, and the product explains this as conceptual similarity rather than exact keyword matching.

In practical terms, you do not need to list every synonym. If your interests include microglia, related papers about neuroinflammation or immune signaling may still be relevant even when the exact word is not prominent in the title. The reverse is also true: a paper can contain your keyword and still not be very useful to you.

Candidate papers and ranking

Weekly recommendations are stored as ranked article rows for each user and week. The current app displays those rows in stored rank order, with a configurable limit on how many articles a user sees. The visible ranking is primarily about relevance to the researcher's stated interests.

Article information used for display includes title, authors, journal, publication date, summary, PubMed identifier where available, journal quartile where available, publication type, and open-access or full-text links where available. Journal-related metadata can be shown to help researchers interpret a result, but it should not be treated as a substitute for reading the paper.

The repository does not expose a public explanation of exact ranking weights, and those details may evolve. NeuroBriefer should therefore be understood as a practical prioritization system, not a fixed or exhaustive bibliographic method.

Weekly briefings and article counts

Weekly briefings are generated for the user's stored interests and then displayed in the dashboard and, when enabled, sent by email. The app lets users choose how many articles to display, up to 50, and how many to include in weekly email. Lite email briefings are described as the top 10 ranked articles.

Favourites, notes, collections, and feedback

Favourites, notes, and collections help researchers keep track of papers after triage. “More like this” and “Less like this” store feedback on recommendation items in the app. These actions help express what was useful or not useful to the researcher.

The current interface says future weekly digests reflect changes to keywords. The code also stores feedback state, but this page does not claim a specific ranking effect or weight for each feedback action beyond what is visible in the application.

Incomplete metadata

Scientific metadata is not always complete. An article may be missing a summary, abstract, journal metric, full-text link, or other field. NeuroBriefer should still be useful when metadata is incomplete, but missing data can affect how easy it is to judge a recommendation.

AI summaries

NeuroBriefer uses AI-generated summaries to help decide what to open first. Summaries are not final interpretations. They can miss nuance, over-compress methods, or fail to capture important limitations. They should point you toward the paper, not replace the paper.

Open access and full text

Where available, NeuroBriefer surfaces open-access and full-text links. This makes it easier to get from discovery to reading. It does not mean every paper is free to read, and it does not remove the need to respect publisher terms, copyright, and institutional access rules.

Major limitations

No ranking system is perfect. Relevant papers can be missed. Irrelevant papers can appear. Keyword choice matters. Article metadata can be incomplete. Summaries can be wrong. Researchers should still check databases directly when coverage matters and should always verify important claims in the original source.