Navegando por Autor "Martins, Gilbert Breves"

Agora exibindo 1 - 3 de 3

Análise de sentimentos no comércio online utilizando bertimbau para o português brasileiro
(2024-03-19) Diniz , Paulo Henrique Barros; Rodrigues, Marlos André Silva; http://lattes.cnpq.br/0650250324042531; Rodrigues, Marlos André Silva; http://lattes.cnpq.br/0650250324042531; Santos, Alyson de Jesus dos; http://lattes.cnpq.br/5998752909180697; Martins, Gilbert Breves
The digitalage,withitsrisinge-commerce,hastransformedthewayconsumersandbusinesses interact. Whileonlineshoppingoffersconvenienceandspeed,ithasalsocreatedaphysical barrier betweenconsumersandbusinesses.Thishasledtoagrowingneedforeffectivedigital communication methods.Onlinereviewshavebecomeanessentialtoolinthisprocess,acting as afeedbackbridgebetweenconsumersandbusinesses.Theopinionssharedonlinecarry considerable weight,directlyinfluencingthepurchasedecisionsofotherconsumers.Moreover, these commentsprovidevaluableinsightsforbusinessestoimprovetheirproductsandservices. Tomitigatethisdistancing,e-commerceplatformsencouragethesharingofopinionsandfeelings of usersaboutproducts,whichsignificantlyimpactsthepurchasingdecisionsofotherconsumers and providesvaluablefeedbackforbusinesses.Facedwiththevastquantityandvarietyof comments, thechallengeofanalyzingthemefficientlyarises.TheBERTimbaumodelemerges as asolution,offeringautomatedsentimentanalysisinBrazilianPortuguese.Experimental results demonstratetheeffectivenessofthemodel,withaccuraciesrangingbetween75%to88% depending onthecategorizationofsentimentsandtreatmentof3-starreviews.
Analysis of the reuse of tires in synthetic grass in Manaus from the perspective of the reverse logistics process
(2024-09-01) Barbosa, Thielsse de Araújo; Tino, Cláudio Fernandes; Martins, Gilbert Breves; Claro Júnior, Luiz Henrique; Oliveira, Elisângela Leitão de; Soares, Márison Luiz; Nascimento-e-Silva, Daniel
Prevenção contra deepfakes: desenvolvimento de um sistema de reconhecimento facial para diferenciar rostos humanos de rostos gerados por IA
(2024-03-19) Silva, Romão Charles Silva e; Santos, Alyson de Jesus dos Santos; http://lattes.cnpq.br/5998752909180697; Santos, Alyson de Jesus dos; http://lattes.cnpq.br/5998752909180697; Martins, Gilbert Breves; http://lattes.cnpq.br/4932200790121123; Rodrigues, Marlos André Silva; http://lattes.cnpq.br/0682962508867807
This work aims to develop a system based on machine learning that differentiates images of people generated by artificial intelligence from real people, a capability that can be very useful for identifying scams that use generated images. The development of the project was done in 3 main steps: data organization, training and testing. The system was entirely made in Google Colab, therefore, it used Python and the main development tool Tensorflow, two models were trained, one of which has almost twice as many images used for training, with the intention of observing the consequences of using a larger set. At the end of the project, the quantitative and qualitative results of the image classification are shown.