Posters.science: Building Open Infrastructure for FAIR and AI-ready Scientific Posters

Description

Abstract

Every year, millions of scientific posters are presented at conferences. They are particularly relevant to biomedical research, where conference posters often represent the earliest public glimpse of ideas and findings that will eventually shape clinical guidelines and drug targets. Yet, most posters disappear by the end of conferences, rolled up under chairs or stuffed into recycling bins. Through our investigation of poster-sharing practices, we found only ~150,000 posters online across 86 platforms, and even those lacked basic metadata, like author identifiers, funding sources, and conference information, making them hard to find and reuse. To address this, we are developing Posters.science, a free and open-source platform that simplifies poster sharing. Users only need to upload a PDF of their poster, review auto-extracted metadata, and publish to a repository of their choice, like Zenodo. We developed a custom LLM pipeline that handles the metadata extraction and generates a poster.json file, which is a structured, machine-readable version of the poster that makes it interoperable and AI-ready. The platform also works as a registry, currently indexing over 1,500 posters from Zenodo and Figshare, with more on the way. For the biomedical community specifically, this could open new possibilities for systematically mining early-stage findings on a disease or gene across thousands of posters. In this talk, we will present our analysis of current poster-sharing practices, introduce the poster.json schema design, and give a live demonstration of Posters.science.