In today’s digital world, bookkeeping and accounting tasks still involve a large amount of paperwork and tedious manual interventions that limit process efficiency and are therefore a barrier to automation.

E-invoicing standards have been developed to reduce manual work; however, most companies, especially the smaller ones, still make exclusive use of either paper, e-mail or pdf invoices, receipts and purchase orders, where the semantic information is not understood by the machine, limiting further processing and automation.

Current off-the-shelf commercial products are either characterised by limited functionality, increasing the incidence of human intervention, or require periodic intervention of highly trained staff to ensure a working system. It is, therefore, desirable to have systems that accurately and automatically speed up the preparation and reconciliation of accounts.

The Automated Document Analysis and Classification for Enhanced Enterprise Efficiency (ADACE3) project has received funding from Xjenza Malta through the FUSION Technology Development Programme to leverage state-of-the-art artificial intelligence algorithms to develop a system that minimises human intervention, making it more attractive to the small company market in terms of cost and ease of use. ADACE3 is a joint effort between the University of Malta and PTL Ltd, a leading technology company, and will result in a minimum viable product that can potentially lead to on-the-field testing and commercialisation.

The system extracts the semantic information with an accuracy in excess of 95 per cent

The process pipeline developed commences with a scanned image of a document, such as a receipt, followed by optical character recognition (OCR) and a multimodal neural-network (NN)-based AI model trained on previously labelled examples of receipts. The system is used to extract the salient information from the digitised text through semantic segmentation and tag such information with various categories, such as date and time, shop address, line items, tax and total.

This information is then provided to the end user, for example the accountant, who would need to reconcile such information with other documents such as invoices and quotations, generate reports, etc. The system extracts the semantic information with an accuracy in excess of 95 per cent. The research team is currently developing a rule-based method to boost the performance even further.

Prof. Ing. Adrian Muscat and Prof. Ing. Gianluca Valentino are the principal Investigator and co-investigator respectively of the ADACE3 project, which is funded through the Xjenza Malta Technology Development Programme. Both are academics at the Department of Communications and Computer Engineering, University of Malta. In addition, the research support officers who contributed to this project are Xandru Mifsud, Leander Grech, Adriana Baldacchino and Said Boumaraf.

Sound Bites

•         A team of researchers has created a digital pathology platform based on artificial intelligence. The platform uses new algorithms developed by the team and enables fully automated analysis of tissue sections from lung cancer patients. The platform makes it possible to analyse digitised tissue samples on the computer for lung tumours more quickly and accurately than before.

•         New research has uncovered an extraordinary mechanism of cell division in Corynebacterium matruchotii, one of the most common bacteria living in dental plaque. The filamentous bacterium does not just divide; it splits into multiple cells at once, a rare process called multiple fission.

For more soundbites, listen to Radio Mocha every Saturday at 7.30pm on Radju Malta and the following Monday at 9pm on Radju Malta 2 https://www.fb.com/RadioMochaMalta/.

DID YOU KNOW?

•         The phrase ‘Mad Hatter’ never appears in the Alice in Wonderland books.

•         Mice have the same number of neck vertebrae as giraffes.

•         Unlike Malta, Suriname is 98 per cent forest.

•         Malta has more than one church per square kilometre.

•         Making commuters stand on both sides of an escalator is more efficient than reserving one side for walking.

For more trivia, see: www.um.edu.mt/think.