The kind of applications and avenues that AI opens up knows no bounds. Machine learning algorithms are now capable of solving human-like tasks involving vision and sound. The computer vision techniques today go beyond images themselves to any type of digital information.

AI, with its unprecedented amount of computing power, now can afford to learn and compute in real-time.

In the case of document processing, machine learning, particularly, can be used for document classification, predictive analytics and decision support for online transactions.

Document processing tools can find their niche in the following applications:

• Valuation of real estate portfolios
• Support documentation for approving loans
• Invoices
• Online verification of identification documents
• Clause detection in legal documents

Automating document processing comes with benefits such as organising information, automatic inspection, navigation and event detection.

Here are a few platforms that provide top document processing AI based services:

Xtracta

Xtracta is a leader in providing automation software fueled by artificial intelligence for document processing. It provides its services to companies like Volvo, where the eDocs is used to cut down on time by 40% consumed in invoice entry.

Xtracta powered services process over 10 million pages per month. It does this with the help of artificial intelligence engine, which requires no manual templates unlike in traditional optical character recognition(OCR) methods.

This AI engine is a “set and forget” engine as it will self-learn new document designs without the need for new templates.

Serimag

Serimag works in parallel with the Barcelona Supercomputing Center (BSC) to classify documents based on neural networks. What makes Serimag standout is its innovation to merge both graphics and text in a document in an integrated manner. And, without the need for parametric coupling modules.

To automate the processing of customers supporting documentation and to standardise criteria, an automatic classification and extraction system was developed by Serimag. This resulted in a reduction in errors and more robust document control systems. And, the company approval cycle has been reduced by hours.

ABBYY FlexiCapture

FlexiCapture platform raises the bar by leveraging machine learning to automatically classify, extract, validate and direct business-critical data, whether it’s from incoming customer communications and operational processes, such as invoices, supporting documents, tax forms, onboarding documents, correspondence, claims, or orders.

Classification technology detects every incoming document type, including images, by using deep learning Convolutional Neural Networks(CNN) and sorts documents by appearance or pattern; and text classification which relies on statistical and semantic text analysis.

And, helps in classifying documents into different types (e.g. bank statement, tax form, contract, invoice, etc.) and variations (e.g. invoices from different vendors) to automatically sort them.

Parascript

Parascript provides computer vision solutions for both image and text classification. They deploy state-of-the-art AI techniques to achieve this. This U.S based company provide the services to the likes of JP Morgan Chase, Lockheed Martin and Siemens.

They employ a topological approach for character recognition using curve tracing powered by neural networks.

Parascript uses computer vision to perform tasks like optical character recognition and handwriting recognition.

Examples of computer vision solutions provided by Parascript include:

• The region of interest location on letters, flats and parcels
• Automatic location detection on envelope images
• Check stock verification and signature verification

Parascript uses convolutional neural networks for deep learning, Hidden Markov Models, Bayesian-based algorithms, support vector machines.

Microblink

Microblink is an R&D company which develops computer vision technology optimised for real-time processing on mobile devices. Advanced neural networks and deep learning techniques are used to provide the most accurate text recognition locally on a mobile device.

Microblink features include:

• Real-time image processing
• Works locally on-device, without an Internet connection
• Supports paper and electronic payment slips in various standards and countries