Framework d’automatització de processos: Classificador de textos fent ús de xarxes neuronals amb la infraestructura de correu electrònic adjacent

Abstract

[en] Based on the need for process automation for cost and time reduction, this project aims to showcase a demo of a possible marketable product that exploits this opportunity by making use of neural networks and automation of Windows process and email services. This project consists of two main parts: the model and the framework. The model: The implemented model is a text classification model that makes use of neural networks. It has been trained with the data provided by the Kofax company, with a total of 700 samples from six different categories; invoices, salary notes, etc. The model has an accuracy of around 90%. It has been benchmarked against other popular classification alternatives used today and optimized as much as possible in order to obtain a satisfactory result within the environment in which it will work. The framework: It consists of a library developed from scratch in order to make use of the model and implement it in a practical, useful and visual way in order to end up with a final product suitable for the market with the title of demo. This framework allows you to deploy an automated email service that works with Outlook. This service consists of receiving, processing, classifying (using the model) and sending data in email format. This library is interchangeable, adaptable and implementable by any user who wants to deploy the service.