Kelvin OCR API

Most Optical Character Recognition (OCR) engines are designed for general purpose use. Further, many of these OCR engines are designed to be used through hosted SaaS APIs. For law firms and legal departments that work with highly-specialized language and highly-sensitive data, these OCR engines are often not a good fit.

The Kelvin OCR API is designed specifically for legal data. Thanks to deep integration with Kelvin NLP and Kelvin Speller, the Kelvin OCR Engine is able to provide legal-specific spelling correction and quality scoring. This means that the Kelvin OCR Engine is able to provide higher-quality text for all downstream tasks like contract review, e-Discovery, due diligence, and contract migration.

The Kelvin OCR API is also designed to be easily deployed on-premises. Thanks to its Docker containerization, it can be deployed on-premises with one line in under 10 minutes. With a private registry, it can even be deployed into airgapped environments without external network access.

Kelvin OCR Engines

The Kelvin OCR API is a modular system that supports multiple OCR engines. By default, Kelvin OCR is configured to use a self-tuning Tesseract workflow, which allows for local, CPU-only processing. For organizations with GPU resources, the OCR API can also also be configured to use PaddleOCR, which may provide better results for some types of documents and languages.

The Kelvin OCR API also supports the use of common cloud OCR engines like AWS Textract, Google Cloud Vision, and Microsoft Azure Form Recognizer and Computer Vision. These cloud OCR engines can be used either as a primary OCR engine or as a fallback OCR engine for documents that are not successfully processed by the primary OCR engine.

Input Formats

The Kelvin OCR API supports the input formats below:


Output Formats

The Kelvin OCR API supports the output formats below: