From the course: Hands-On AI: Knowledge Graphs for Generative AI Use Cases

Unlock the full course today

Join today to access over 24,900 courses taught by industry experts.

Disputed fact verification

Disputed fact verification

From the course: Hands-On AI: Knowledge Graphs for Generative AI Use Cases

Start my 1-month free trial Buy for my team

Disputed fact verification

“

- [Presenter] It's good to note here that with fact verification, you don't have to fact verify every statement in your graph. As we go through this, it might seem a little daunting to think that you have to fact verify every statement ever created from your data. Again, this is something that you really want to focus on so that you reduce those hallucinations in your LLM. This can start with one area of your product line or organization that needs higher accuracy or something that already has hallucinations, something that's more time sensitive, or things that are just business critical. You can use survey tools like Mechanical Turk or your own internal surveys to solicit feedback on which statements are accurate from your human reviewers. Sometimes there just won't be a winner, this happens for things that are opinion-based, they don't have enough evidence yet to be corroborated, they are truly a valid dispute or they are just bad data. You can choose if you want to retain the bad…

Contents