READ · Document Extraction
Extract your own fields
Describe the fields you need as a JSON Schema and run them across any set of documents.
Built-in extraction covers the fields most land teams track. When you need different ones, describe them as a JSON Schema and run them over any of your documents.
1. Describe the fields
Each property is a field. Its description tells Rowland what to look for, in plain words:
{
"type": "object",
"properties": {
"lessor": { "type": "string", "description": "Every lessor named" },
"royalty": {
"type": "number",
"description": "Landowner royalty as a decimal, e.g. 0.1875"
},
"pugh_clause": {
"type": "boolean",
"description": "Whether undeveloped depths or acreage are released at the end of the primary term"
}
}
}2. One record per document, or per item
per: "document"(the default): one record per document, such as one per lease.per: "item": one record per item found, such as every tract in an assignment or every well on a statement. Say what an item is withitem, such as"tract".
3. Start it
Start an extraction with your schema and the documents to read. Add instructions for anything the schema cannot say, such as which royalty to use when a lease states two. You get an id starting ext_.
4. Collect the records
Get the extraction until its status is completed, or wait for the extraction.completed webhook. Each record carries the document it came from, its values and a citation for every value.
Tips
- Write descriptions like instructions to a new analyst. Units, formats and edge cases belong there.
- Try a handful of documents before running hundreds.
