LLM for automated metadata generation?

cptnstormfield
Is there a good solution to use an LLM to extract metadata from PDFs and add that metadata to Zotero? I have many PDFs in my library that are scans of old journals/newspapers, etc. I've been manually inputting the metadata, which is tedious.

FWIW, ChatGPT is surprisingly adept at extracting metadata using the web interface. I just need to automate it.
  • martynas_b
    Can you show examples of those old scanned journals and newspapers?

    Maybe send some to support@zotero.org and include a link to this thread.
  • cptnstormfield
    I am happy to share examples, but I don't know how enlightening that will be. They are (hobby) journals from the 1960s. Just PDFs, sometimes OCRed, but no embedded metadata.
  • tim820
    edited 19 days ago
    If you have references to those works in the form of a text bibliography, you could import them using the Anystyle web app.
    https://anystyle.io/

    ChatGPT also does a reasonable job of doing the same thing - converting text bibliographies to import formats like bibtex. If you only have more basic information on each item, you could perhaps try asking ChatGPT to generate a bibliographic entry for that item, which could then be converted/imported.
    https://threadreaderapp.com/thread/1753785908266963066.html?utm_campaign=topunroll

    Even writing a bibliography manually and then importing it might be easier that entering metadata manually in each individual Zotero field for lots of items.

  • cptnstormfield
    Thanks, @tim820. Good suggestions.
Sign In or Register to comment.