Mineração de textos para o tratamento automático em sistemas de atendimento ao usuário

Abstract

The explosion of new forms of communication between companies and new opportunities and means for companies to take advantage of this interaction. The way customers interact with companies has evolved in the recent years due to the increase in mobile devices and Internet access: clients who traditionally requested phone service migrated to electronic means of service, whether via smartphone app's or via customer service portals. As a result of this technological transformation of the communication medium, text mining has become an attractive form for companies to extract new knowledge from the register of interactions carried out by customers. Within this context, the telecommunications environment provides the inputs for conducting experiments due to the large volume of data generated daily in customer service systems. This job aims to analyze if the use of text mining increases the accuracy of data mining models in applications involving free texts. For this purpose, an application is developed that aims to identify clients likely to leave internal service environments (CRM) and migrate to regulatory agencies in the telecommunications sector [Baeza, Ricardo e Berthier ,1999]. Also addressed are the main problems encountered in text mining applications. Finally, the results of the application of classification algorithms on different data sets are presented for the evaluation of the improvement obtained with the inclusion of text mining for this type of application. The results obtained show a consolidated gain in the improvement of the acuraria in the order of 32%, making the mining of texts a useful tool for this type of problem.