• Funderpants @lemmy.ca
    link
    fedilink
    English
    arrow-up
    8
    ·
    edit-2
    6 months ago

    We’re experimenting with retrieval augmented generation for early inquiries right now. We get hundreds of inquiries that could be answered by looking at the website/docs and Q&A models with extractive or abstract approaches, or newer generative approaches are good at handling them.

    Looked at four models last week, 2 vendors and 2 open source solutions, it’s very promising. Very high accuracy with extractive approaches to simple queries, an email answering bot that links to our live website, along with an offer to talk to a real person could help us out a lot.