Document Information Extraction

Parse source material, extract structured fields from a schema, and save the results. Use it for contracts, receipts, and scanned documents.

Prepare your resources

Prepare a contract, receipt, or scan with clearly defined fields, an extraction schema, and an output volume. Validate one document before expanding the input set. Use Extract Information to Table for database writes.

Use the template

On the workflow creation page, start from a template, search for Document Information Extraction, and load it onto the canvas.

Complete the required fields in the parameter form.

Setting

Purpose

Source files

PDFs, scans, images, or folders to extract from

Parsing tier

Controls parsing speed, quality, and cost

Extraction schema

JSON Schema defining fields to extract from parsed content

Output location

Catalog location for extraction results

Save the parameters, then start a manual run. The system saves the current workflow before submitting execution.

Complete processing flow

Arrows show execution order; configured bindings supply each step’s input.

Document Information Extraction

Define the information to extract

Use JSON Schema for field names, types, and descriptions. This example extracts a contract number and allows null when it is absent.

{"type":"object","properties":{"contract_number":{"type":["string","null"],"description":"Contract number; return null if absent"}},"required":["contract_number"]}

Explain how to distinguish the contract number from an order number or filename. Missing-value instructions must agree with schema types.

Setting

Adjustment

Parsing tier

Standard by default; try Enhanced for scans or complex tables and inspect text first

Extraction instructions

Explain meaning, format, and missing-value handling

Layout-sensitive fields

Name fields that require page-position context, such as supplier and recipient

Output location

Volume or directory receiving results

The template uses n_to_1, combining the input into one result. Multiple files do not automatically produce separate records. For per-file results, change extraction mode and downstream saving bindings; see Information Extraction.

Inspect the run

  1. Confirm the selected files in reading results.

  2. Compare parsed text with the original, especially identifiers, dates, and amounts.

  3. Inspect result for field values, types, and missing values.

  4. Check available sources_tracking information.

  5. Open the destination and verify returned files. The template saves extracted documents and results with ZIP configuration.

Parsing supplies documents; extraction produces fields. If text is correct but fields are wrong, refine instructions. If content is missing, resolve parsing first.

Example: extract a contract number

Select one contract, use the schema above, and request its formal number from the body. Compare the result character by character with the original and inspect the saved file. Attachments belonging to the same contract can contribute to one combined extraction.

Common problems

Symptom

Check and action

Empty field

Check whether the information exists and survives parsing

Wrong identifier selected

Distinguish contract, order, and attachment numbers in descriptions

Null violates the schema

Align field types with missing-value instructions

Several files become one record

The template combines inputs; per-file processing needs mode and saving changes

Table records are needed

Use Extract Information to Table and verify column structure
Last updated on