Notion: https://elianna.notion.site/elianna/Automating-Metadata-a8acd0a54e05497dad0faf3ada5f1708

The goal of this project is to ignore the current metadata system and start over. We want to know

• What metadata is necessary for science?
  • Resources
    • elicit.org - This search engine is already scraping metadata through GPT-3 including customization of which data you the user wants. Sometimes it doesn’t return results.
      • I’m not sure if they use GPT for the base metadata (authors, etc) or if they use Semantic Scholar’s API.
    • https://www.researchobject.org/ro-crate/1.1/ -> a metadata standard.
      • https://research.manchester.ac.uk/en/publications/ro-crate-metadata-specification-111 - the documentation for the project above.
    • Schema.org
    • FAIR data standards
    • Data standards: (
      1. Data origin: experimental, observational, raw or derived, physical collections, models, images, etc.
      2. Data type: integer, Boolean, character, floating point, etc.
      3. Instrument(s) used
      4. Data acquisition details: sensor deployment methods, experimental design, sensor calibration methods, etc.
      5. File type: CSV, mat, xlsx, tiff, HDF, NetCDF, etc.
      6. Data processing methods, software used
      7. Data processing scripts or codes
      8. Dataset parameter list, including
         • Variable names
         • Description of each variable
         • Units
  • Elaboration
    • We should explore this at a granular level at the research object (instead of just research node). What is the metadata for code, pdfs, etc?
    • We’re in the exploratory phase -> we want to see what data is in these papers. What can we get out of this?
    • Error in GPT isn’t a super huge problem at the moment -
      • We’re trying to find ways to make things better!
• How can we eliminate the need for scientists to enter that metadata?
  • Hypothesis: GPT!
  • Other?
• What else?

What we have

• A section for metadata on the node platform.
  • Build out/explore what the Nodes team has discovered through scraping PDFs
  • later on down the road -> integrated into search. But shouldn't think of this as integrating into the nodes platform.
• Emerging Problem statements - (To be edited!)
  • There is no understanding of what metadata is prevalent across all research. There is no standard for that.
    • Connected to ontology.
  • Longer term problem statement -> researchers don't have the knowledge processes/workflows to correctly enter metadata in a standardized format.
    • metadata that is unstandardized and it is difficult to read and use.
    • Difficult to seamlessly tie one piece of research to another (what are the relations between research) ununified scientific record.
    • How might we unify the scientific record w metadata standards that can be accessed by different platforms/gateways so that they can use the research in any way they want.