Extract tables with Collections
Use Collections to create structured data
Collections are a powerful property type used to extract structured, tabular data. For example, you can use a Collection to extract individual line items from an invoice, a table from a legal contract, or a breakdown of product lines from a pitch deck.
These item-level extractions can then serve as inputs for other properties, allowing for deeper analysis and automation.
Here's how you can create your own Collections:
To create a collection:
- Add a new property with the Collection type

- Select an AI Tool and an Input for the property
- Add the properties (columns) that you'd like to extract, and any context on what these properties represent to the prompt

- Click on Review to review the extracted fields against ground truth

Any extracted fields can then be used as inputs into other properties. For example, after extracting the cost for every line item from each receipt, we can then sum these with the Number property to create a total cost per receipt.

Updated about 10 hours ago