Skip to content
getnextpdf.com

Enterprise edition

Intelligence — Deep Reference

This deep reference documents key-value typing and schema validation, table grid synthesis, and the searchable-overlay orchestration boundary.

This capability ships in NextPDF Enterprise (nextpdf/enterprise) and activates with an Enterprise-tier license envelope. A deployment without that entitlement does not load the capability’s classes. Compare editions and get a license.

StructuredExtractor::extract types raw key-value input. When a schema is supplied, only pairs whose key matches a schema field are kept; pairs without an explicit confidence receive a fixed default. validateSchema returns the required field names that were not found (empty means compliant). This step types and validates already-extracted data; it does not recognize text and assigns no computed precision.

TableExtractor::extract builds structured tables from raw row grids. The column count is the widest row; short rows are padded with empty cells. Each cell receives a synthesized bounding box from a uniform subdivision of a normalized page area and a fixed default confidence. A table with no rows or no columns is skipped. The bounding boxes are a grid synthesis, not a measured layout.

SearchableOverlay::generate validates the PDF header, bounds input size and page count (failing closed on overflow), detects pages lacking a usable text layer, drives the configured OCR backend, and returns per-page word counts and an average confidence. A page with a usable native text layer is skipped by default. Invisible OCR text relies on a text rendering mode that neither fills nor strokes the glyphs (ISO 32000-2:2020 §9.3.3); when the source is tagged, the structure tree maps content to a reading order (§14.7) and table structure elements describe rows and cells (§14.8). The actual rasterization and invisible-text injection are delegated to a separate sidecar process; the PHP surface orchestrates and returns result metadata. The overlay output is a derived document — any existing signature is invalidated and compliance must be re-validated. Confidence is passthrough or a fixed default; the surface asserts no OCR accuracy or extraction recall.

NextPDF\Enterprise\Intelligence\StructuredExtractor, NextPDF\Enterprise\Intelligence\ExtractionSchema, NextPDF\Enterprise\Intelligence\SchemaField, NextPDF\Enterprise\Intelligence\KeyValuePair, NextPDF\Enterprise\Intelligence\TableExtractor, NextPDF\Enterprise\Intelligence\TableResult, NextPDF\Enterprise\Intelligence\TableCell, NextPDF\Enterprise\Intelligence\SearchableOverlay, NextPDF\Enterprise\Intelligence\OverlayConfig, NextPDF\Enterprise\Intelligence\SearchableOverlayResult, NextPDF\Enterprise\Intelligence\PageOverlayInfo, NextPDF\Enterprise\Intelligence\ExtractionConfig, and the ExtractionStrategy / OverlayQuality enums. The OCR backend is an injected contract supplied by the deployment. Signatures are listed on the public page.

The overlay mechanism maps to ISO 32000-2:2020 §9.3.3 (text rendering mode) and, for tagged sources, §14.7–§14.8 (structure tree and table elements). The structuring steps consume already-extracted data; quality is bounded by the upstream extractor and the OCR backend.

  • Synthesized table bounding boxes are a uniform grid subdivision, not a measured layout.
  • When the OCR backend is unavailable, affected pages contribute zero words rather than failing the whole run silently; check the per-page result.
  • The overlay output is a derived document; re-validate signatures and compliance status.
  • No cryptographic operation occurs in this module, so there is no FIPS-mode-specific behavior.

This page documents externally observable behavior and the supported public API surface only. Internal namespace paths, helper classes, mechanism tables, runbook filenames, and ticket prefixes are out of scope.

NextPDF Core (Apache-2.0) has no searchable-overlay orchestration and no schema-validated structuring surface — none; this capability has no Core-tier equivalent.

NextPDF Pro ships structural AST-driven extraction over an already-parsed document; it does not provide schema-validated key-value structuring, table reconstruction from raw grids, or OCR-backed searchable-overlay orchestration. Those ship in the nextpdf/enterprise package only.

Key-value typing, schema validation, table grid synthesis, and overlay orchestration are described at the behavior level. The actual rasterization and invisible-text injection run in a separate sidecar process; the sidecar internals, the OCR model, and any internal orchestration detail are out of scope and are not reproduced here. The OCR backend is an injected contract supplied by the deployment.

The deployment supplies and operates the OCR backend and chooses where OCR is computed — and therefore where document images and recognized text are computed and stored. The surface bounds input size and page count and fails closed on overflow. NextPDF Enterprise orchestrates the workflow and returns result metadata; it does not embed an OCR model or guarantee recognition accuracy or extraction recall.

No export-control restriction applies to this surface. The surface asserts no OCR-accuracy or extraction-recall guarantee and no regulatory-compliance status. This reference is not a legal opinion.