Modelagem e incorporação de um banco de dados orientado a grafos ao Chatbot LIA

Abstract

Considering the advances in technology, there was an increase in the number of applications that serve the diverse areas of computing, from CRM (Customer Relationship Management) systems to networks that make use of the Internet of Things. These applications usually need a database system to save and retrieve the necessary information quickly and safely. Although all existing databases fulfill this purpose, the various database models differ due to factors whose degree of importance depends on the application. These factors include, but are not limited to, the amount of data to be stored, how this data should be organized, and the reading and writing time made by the database. Taking into account chatbot applications, it is known that there is a continuous search for methods that make the bot conduct more fluid conversations, be agile in decision-making and provide a quality service. One of the main mechanisms to achieve these objectives is to add a knowledge base so that the chatbot has access to all the resources necessary to make autonomous and intelligent decisions. Thus, this work is based on the modeling of a graph database that acts as a knowledge base for the virtual assistant LIA, which was created to serve the real estate industry. The first step consisted of conducting a systematic review to search for related works to this subject, that is, the integration of graph databases in chatbots. Then, market research has been made to select the most suitable graph database for the chatbot LIA considering several pre-established objective criteria. Finally, the advantages of applying this new database model were compared to the original version of the chatbot. It was found that Grakn was the most suitable among all the graph databases analyzed and that it allowed the modeling of a knowledge base richer and more flexible than the current version of the LIA.