Text Recognition

The Text Recognition module provides raw OCR (Optical Character Recognition) capabilities independent of pipeline-based extraction.

Overview

While the standard extraction pipeline performs intelligent, structured data extraction, the Text Recognition module offers a lower-level interface for obtaining raw text content from documents and images.

Capabilities

• Full-page text extraction with position coordinates
• Multi-language text detection and recognition
• Handwritten text recognition
• Table structure detection and text extraction
• Line-by-line and word-by-word segmentation

When to Use

• When you need raw text without structured field extraction
• For documents not covered by existing pipeline configurations
• When building custom post-processing logic on top of OCR output
• For text search and indexing workflows

Output Format

The text recognition response includes:

• pages — Array of page objects with extracted text
• lines — Individual text lines with bounding box coordinates
• words — Word-level segmentation with confidence scores
• language — Detected language(s) of the document

ℹ

Note: For structured data extraction, use the standard pipeline-based extraction endpoint instead.