• pearsaltchocolatebar@discuss.online
    link
    fedilink
    arrow-up
    8
    arrow-down
    1
    ·
    6 months ago

    You can’t. An NDA prohibits me from saying how I know this, but I know for a fact that advanced and specifically trained FMs still struggle to accurately parse resumes, even with several million dollars devoted to the project.

    • MacN'Cheezus@lemmy.today
      link
      fedilink
      English
      arrow-up
      4
      arrow-down
      1
      ·
      6 months ago

      Really? That’s quite surprising. I understand that it’s not trivial to algorithmically parse a resume formatted for human consumption even though it’s a somewhat structured format, just because the formatting can vary quite a lot, but there’s only so many different types of information on there, and little of it has any overlap in terms of how it could be categorized, so I would think an AI should be quite effective at picking it up.

      Then again, I’m not an AI expert and I certainly haven’t attempted to do anything like this.

      • pearsaltchocolatebar@discuss.online
        link
        fedilink
        arrow-up
        3
        ·
        6 months ago

        The models can do alright for very simple resumes, but once you start getting into multiple page ones it gets messy.

        While resumes follow a pattern that we’re able to easily recognize, there’s so much variation that even with AI you have to add in a ton of heuristics to control for hallucinations.