Our Approach

Humans & machines work better together

Rossum’s unique cognitive data capture technology mimics the human mind when detecting and capturing document data. It then gives human operators intuitive assistance, enabling them to validate and correct captured data in seconds.

Our cloud-based technology ensures best-in-class security and scalability, as well as full 24/7 access on any device.

Humans and machines work better together.

Precise Data Capture Is a Must

Invoice data capture is inherently connected to numerous business functions. The receipt of an invoice triggers a series of processes that have specific data requirements. To process an invoice, several data fields must be localized and data must be extracted from those fields.

Cognitive data capture uses advanced machine learning to accelerate these steps and ensure accurate results.

Precise data capture is a must.

How Does Cognitive Data Capture Work?

Cognitive data capture is a two-step procedure: data extraction followed by validation and correction. This delivers a set of data values paired with metadata information for use in related business processes.

Much like the way humans scan documents, Rossum uses a unique type of spatial OCR to scan invoices to understand their structures, patterns, and potential meanings. The platform’s neural networks then generate candidates, choosing the most suitable ones based on their confidence scores.

Following this, Rossum reads the data carefully, capturing each field’s content. As the precision of character location increases, the application uses a proprietary OCR solution to transcribe the final text strings, assigning each one a confidence score.

Cognitive data capture consists of initial data extraction followed by validation and final extraction.

Human in the loop

In the validation and correction stage, Rossum automatically prompts the user to inspect empty fields and review data with low confidence scores. The platform then directs the user to relevant areas of the document so they can add or change data in a few clicks and/or keystrokes. This human feedback is an integral part of the technology, enabling it to learn and become more accurate.

Rossum enables human operators to correct and validate captured data interactively.

Self-Learning AI

Rossum’s out-of-the-box AI comes pretrained to understand hundreds of thousands of invoice templates, ensuring highly data extraction from day 1. The human feedback component enables every business to train their own dedicated version of Rossum’s AI to meet their specific requirements.

Self-learning AI gives any business the opportunity to train their own dedicated version of Rossum’s AI.

Cognitive Is Better than Manual

Most businesses are still using manual data capture to process invoices. Cognitive data capture is faster, more accurate, more efficient, and less costly, and it frees human operators to play value-generating roles. Also, as a cloud-based solution, Rossum can be set up and updated in no time.

Cognitive data capture is faster, more accurate, and more cost-effective than manual data entry and template-based OCR data extraction.

Cognitive Beats Traditional OCR

Traditional OCR systems solve some manual data capture issues, but they also create new ones. Human operators must write rules and templates for every invoice layout, making maintenance a neverending chore.

Because it is document-agnostic and flexible, cognitive data capture technology solves these problems, delivering increasingly accurate results with continued use.

Unlike traditional OCR data extraction, cognitive data capture does not require templates.

Read more about Rossum technology in our
blog and in academic papers