MegaBot: uma proposta para aprimorar o atendimento das ouvidorias públicas do estado do Amazonas

Abstract

The chatbot tool, the result of artificial intelligence, has been widely implemented in public bodies and entities in order to improve interactions, and has proven to be of great value in various services. In view of this, with the use of the chatbot tool, it will be possible to carry out automatic queries of information in its own database, eliminating direct human interference. In this sense, the objective of the work was to develop a chatbot-type system that makes use of natural language processing, adapted to the Portuguese language, using an open source library, so that each public ombudsman in the State of Amazonas that is interested in the implementation of the tool can manage it freely. As for the applied methodology, bibliographical and practical characteristics were chosen, considering that it was the development of an intelligent agent (chatbot) demanding both characteristics. To this end, a systematic literature review was carried out, as well as chatbot training through 94 questions and answers placed in the chatbot's own database using Natural Language Processing (NLP) and Machine Learning technologies. In this way, intentions within a specific context were detected, training and tests were carried out to learn the chatbot, with the aim of developing a system that learned more and more from interactions with users. Finally, it was possible to obtain a very positive result in the training sessions, as well as in the tests performed. As a result, the chatbot developed was called MegaBot and submitted to an assertiveness assessment through 156 questions, equivalent to 780 tests. For these tests, the system registered 233 errors and 547 correct answers, thus achieving a result of 70.13% of correct answers, considered a good assertiveness index. Therefore, the work reached its objective with the development of a complete system of interaction with the user, for the improvement of the service of the users of the public ombudsman.