Document classi铿乧ation using deep learning

Abstract

This study tackles the classi铿乧ation of business documents into six pre-de铿乶ed classes, (invoices, receipts, delivery notes, purchase orders, quotations and others). Three machine learning models, in increasing complexity, are proposed, implemented and compared. The models comprise a term frequency based classi铿乪r, a TF-IDF based Multinomial Naive Bayes classi铿乪r and a TF-IDF based Arti铿乧ial Neural Network classi铿乪r. The models are trained and tested using a synthetic business document dataset, which was created as part of this project. The Neural Network classi铿乪r obtained the highest overall classi铿乧ation accuracy, over 97.7%, and outperformed the other two models by at least 5% points. The results show how this task bene铿乼s from the implementation of deep learning.