Identifying and extracting handwritten information is still a challenge in the scanning process. The reasons for that are plenty: original documents may have poor quality as paper deteriorates easily; notes may have been written on the fly; signatures are almost always unreadable (not to mention that half of the population writes "1" as if it were "7" and the other half writes "7" as if it were "4"!) All jokes aside, let’s face it – we don't concentrate on having good handwriting when filling in forms, and most of the time we don't write within the text fields on the document.
Today you have a choice between 3 different technologies for recognising and extracting handwritten information:
- ICR – Intelligent Character Recognition – low recognition accuracy, and therefore high validation efforts
- IDR - Intelligent Document Recognition – medium recognition accuracy, and medium validation efforts
- DNN - Deep Neural Networks – high recognition accuracy resulting in low validation efforts
When properly trained for specific document formats, Deep Neural Network technology - including categories like Convolutional Networks, Recurrent Networks, Attention Networks, Differentiable Networks, etc. - can reach up to 99% accuracy at field level. A relevant example is Moonoia processing handwritten medical attestations for Belgium health insurance companies:
- Approximately 8.000.000 claims processed/year for a single customer
- Average recognition rate for most fields: >98%
- Straight through processed forms: > 70%
- Average turn-around time : < 24 hours
Interested in learning more? Join the upcoming Moonoia session at the AIIM UK Forum “Using Deep Learning to Achieve Real Straight Through Processing” or visit www.moonoia.com to learn about how AI-powered data extraction technology can transform information processing within your company.