Nutrient Launches Document Data Extraction API With Per-Field Citations and Confidence

Executive Briefing
“Nutrient has released a document data extraction API that maps extracted fields to a JSON...”
Key Takeaways
- 1.Nutrient's data extraction API parses PDFs, scans, images, and Office files into spatial JSON or Markdown.
- 2.Each extracted value is anchored to a page via a bounding box, a match label, and a confidence score.
- 3.In Nutrient's internal WholeDoc Bench, the agentic configuration scored an overall 86.2 at a cost of 1.8 cents per page.
Nutrient has released a document data extraction API that maps extracted fields to a JSON schema with page-anchored citations and confidence scores. The platform supports multiple processing modes including structure, understand, and agentic workflows.
Product Features & Key Highlights
Nutrient's data extraction API parses PDFs, scans, images, and Office files into spatial JSON or Markdown.
Each extracted value is anchored to a page via a bounding box, a match label, and a confidence score.
In Nutrient's internal WholeDoc Bench, the agentic configuration scored an overall 86.2 at a cost of 1.8 cents per page.
Practical Assessment & Evaluation
Integrating match labels like id match alongside confidence scoring helps developers build automated validation gates before accepting data extracted from unstructured documents.
Supporting multiple modes ranging from basic text parsing to vision language models allows teams to balance cost and accuracy based on document complexity.
Pricing, Availability & Next Steps
Real-world performance on independent public datasets outside Nutrient's internal WholeDoc Bench evaluation remains unverified.
The exact accuracy when handling cross-page spanning tables in complex enterprise environments is not fully detailed.







