Ontologias e processamento de linguagem natural na publicação e acesso de dados para carta de serviços do Ceará

Abstract

This project seeks to develop a database on service letters, implement a data ontology for publishing open data for the Ceará service letter, integrating information from adverse status letters linked to a chatbot to facilitate access information for cities. The work implemented a comparison of data on state service charters, aiming at a generalization of the terms and with the support of technologies such as Protegé and the python language, the rasa framework and its libraries for the creation of the ontology and chatbot . The T-Box and A-Box concepts were defined to fill the ontology. When filling in the ontology, a small comparison was made on two equivalent services from different service letters, pointing out their differences in certain fields. 1400 test questions were applied and the results obtained an accuracy of 0.340 at the conversation level, 0.477 at the entity level and 0.849 at the intention level for the associations made by the chatbot on its trained database.