Mobile acquisition of accounting documents and processing of raw extracted data

Abstract

Over the past decade, the job of accountant has evolved tremendously with the rise of the Internet. Everyone is always looking for ways to do more in less time and in the case of accountants, a lot of energy is still wasted in recovering invoices from their clients and encoding them manually into their system. The goal of this thesis is to provide an efficient way for the client to scan documents with a mobile application, and seamlessly send them for treatment. For each received invoice, the server performs an automatic pre-encoding of all fields by using OCR techniques for text extraction and other methods for detection and selection of the right fields inside the document. Then, the output can be integrated into the accountant's program. The resulting system is time-saving for both parties. Besides, it is capable of auto-improving with time, allowing for increasingly accurate pre-encoded documents to be received by the accountant. Also, it allows him to be frequently provided with single invoices rather than to be rarely supplied with an important amount of them. While some fields remain to be included into the detection, most of them are already fully integrated. The application is aimed for Belgian French-speaking users, but could be adjusted to other countries and languages in the near future.