New Patent 10,936,863

We are very proud and excited to report that our latest patent was published. 

This marks the innoVAT\GSTive work being done in our machine learning and data science group to advance the public knowledge of information extraction from document images. 
The technology described in the patent is central in our automated end-to-end VAT\GST reclaim product, and drives our business.

What the patent is all about 

This patent describes how convolutional neural networks can be used to extract information from invoice and receipt images. 
By giving the convolutional network visual, geometric, linguistic and semantic information, we teach it to find a representation for key information in the invoice: Total amount, VAT\GST amount, and so on.

The network is tasked with finding in the image the positions of these fields, by marking the pixels that cover them.
We use an inverse-softmax loss function to allow the network to possibly find multiple locations for each field.

This deep learning technology has shown a very big increase in performance over existing and off-the-shelf information extraction solutions.

The value to our partners and customers.
Cutting-edge technology for extracting information from invoices makes our product for VAT\GST reclaim faster, more economical and higher performance.

Our clients and partners benefit directly from the technology by receiving their claims with quicker turnaround times and with more invoices processed compared to other solutions out there.

We’re looking forward to soon share more information of patented technology from our research team at WAY2VAT, and the way it revolutionizes the VAT\GST reclaim process.