What is schema-drivenextraction?
Schema-driven extraction means defining the data you need once, as a schema, and extracting it from any document layout by meaning: no templates, no per-type pre-training.
Describe the data, not the document.
In schema-driven extraction you describe the output: which fields, which types, which structure. The extraction finds those fields in the document by meaning, wherever the layout puts them. The alternative approaches describe the input instead: templates map pixel positions per layout, and pre-trained models fix a catalog of supported document types.
The difference shows at the edges. A new supplier, a redesigned payslip or a document type invented by your own industry needs no new setup: if the schema describes it, it extracts. Maintenance shifts from keeping templates alive to occasionally refining a schema.
One definition
A schema describes fields, types and structure once, for all layouts of that document class.
Extraction by meaning
Values are found semantically: gross pay is found as a concept, not at a coordinate.
No catalog limits
No pre-trained type list: any document you can describe becomes extractable.
The core of MiruIQ.
Schema-driven extraction is MiruIQ's founding design choice: schemas are created in the Studio or generated by the wizard, versioned, and reused across pipelines, with validation and human review built around them.
Schema-driven extraction questions
What evaluators ask about the schema approach.
A template says where values sit on a specific layout, so every new layout needs a new template and every layout change breaks one. A schema says what values mean, so layout variance is absorbed by the extraction itself. In template systems maintenance grows with your supplier count; in schema systems it does not.
In MiruIQ, visually in the Studio, assisted by a wizard, or as JSON Schema via the API. A schema lists fields with names, types and descriptions, including nested structures for line items and tables. Most teams start from a sample document and refine after the first trial runs.
See it on your own documents.
The 14-day free trial covers 200 documents: enough to define your schemas, run real files and measure the results before anyone calls you.