Does anyone have any tips on how to shape data in our portals to work better with the AI tools? The concern from our leadership is someone will ask a question and if the data contains too many filterable items it could return an incorrect result.
Is there any guidelines on how to better shape the data to make it easier for AI to understand and provide results?
Hi Danny,
I'm sharing with you the guidelines we've adopted for using datasets in our AI tools. The two main points are:
Complete Metadata: To ensure datasets are easily searchable and discoverable, it's crucial that metadata such as the title, description, and keywords are fully completed. Keywords that are easily understood by LLMs are particularly important.
Detailed Schema: It is also essential to define the dataset's schema by providing a clear name and description for each field. For fields with a fixed set of options, the list of all possible values should also be described.
In any case, I'm very interested in this exchange.
No account yet? Create an account
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.
