Overview

The product “Training as a part of the workflow” is a capability of the CMR+ platform from AntWorks. It embeds training of machine-learning (ML) models into the document-processing workflow, so that models continually learn, adapt and improve accuracy as they handle more documents. Users supply corrections or feedback during processing (human-in-the-loop), enabling active-learning, iterative retraining, and improved performance over time, aligned with evolving document patterns and formats.

Features and Capabilities

• Feedback & Correction Loop: Users provide corrections/annotations when document processing errors occur; these feed back into the ML models to improve accuracy.
• Active Learning Selection: The system intelligently selects documents that are uncertain or challenging for human review, optimizing training effort and efficiency.
• Iterative Retraining: Training is integrated into the workflow so models continuously retrain as new labeled data becomes available, staying aligned with changing document types and industry terminology.
• Multi-Format & Document-Type Support: Works across images, PDFs, MS Word, handwritten inputs, unstructured/structured data sources; supports ingestion of diverse filetypes and languages.
• Human-in-the-Loop Verification & Training: Users (including citizen developers) can verify outputs and provide training feedback, bridging automation and human oversight.
• Integration with Workflow & Automation Systems: Training module embeds within document-processing workflows, enabling automation supervision, exception handling, micro-services and RPA as part of end-to-end process.