OCR
Extract text accurately from any source — printed, handwritten, or scanned
Prism’s OCR tests ensure reliable text extraction, layout integrity, and language consistency across diverse documents and image conditions.
Key Capabilities
Catch Reading Errors
Detect missed or misread characters across printed and handwritten text.
Fix Extraction Inconsistencies
Identify inconsistencies in entity extraction, layout, or formatting.
Maintain Text Integrity
Compare OCR outputs across model versions to maintain text integrity.
Tests
An overview of all tests
Outlier Detection
Identify scenarios in the field data which are drastically different (out-of-distribution) with respect to the training dataset.
This automated test helps users access if the data in the production setting has shifted and the model needs to be retrained.
Missing Detections
Identify label drift among the model predictions
This automated test helps access if the label distribution in the production setting has shifted and the model needs to be retrained.
