• WVbrU@lemmy.world
      link
      fedilink
      arrow-up
      1
      ·
      29 days ago

      I don’t really understand (or agree with) their proximity calculation. For example:

      Tap for spoiler

      It said “shell” was tepid when the answer was “seashell”. However, “driftwood”, “pinecone”, and “doily” are some of the closest proximity words…

  • Theroddd@lemmy.mlOP
    link
    fedilink
    arrow-up
    1
    ·
    edit-2
    29 days ago

    I don’t fully understand the way it works myself.

    I read though the description and admit that I don’t really get what the AI is doing.

    I agree though, today “clam” was much furthur away that “marble”. I think maybe seashell and marble are both decorative and likely to appear near each other in an article. Clam and seashell, while one can be part of the other, and are closer in my mind, are not as linked in articles.

    From your example, driftwood, pinecone, doily, are decorative items, so if it’s pulling from interior design topics, I can see how they are likely to be related.