Classificação de texto : multinomial x bernoulli utilizando reviews de E-Commerces

Abstract

With the popularization of the internet many services that previously could only be accessed physically came to offer its digital version, one of these services is the e-commerce which is the representation of the store virtually. As this relationship between customer and service provider is only virtual, some customers have sent reviews to companies in order to explain their satisfaction or dissatisfaction with the service. In this way, some companies like ReclameAqui and Trustvox started to assist e-commerces and clients to facilitate communication and consequently to increase the quality of the products and services offered. The following work aims to develop a predictive model capable of classifying, given a review, which category it refers to, the labels used are: information, characteristic, price and delivery. For this, the pre-processing and cleaning step in the used base was performed, which was provided by the company Trustvox, then the training and testing of the Multinomial and Bernoulli models were performed and finally the result of the comparison of the performance of the models for each label .