Desenvolvimento de um Modelo Ensemble para Responder Questões via Bancos de Dados Estruturados (KB-QA)

Abstract

The virtual sales market (E-commerce) has expanded a lot nowadays due to the ease and practicality provided by this way of purchase, and the fact that technologies are becoming increasingly accessible. With this, the implementation of virtual assistants by companies can add benefits for both sides of the negotiation (consumers-companies), since the use of virtual assistants can allow the automation of tasks (involving means of communication), thus accelerating solving problems and increasing productivity for customer service, in addition to being able to provide personalized experiences tailored to each customer. In this work we discuss some models and strategies that are relevant to the topic of Knowledge Base Question Answering (KB-QA), and means of developing a KB-QA model capable of answering user questions provided in natural language, based on information contained in knowledge bases (KBs) that can be implemented as part of a virtual assistant. In order to develop a superior KB-QA model via Ensemble by majority vote, the search, analysis and selection of KB-QA models are carried out for the Ensemble’s composition. For the analysis of the KB-QA models, the WebQuestionsSP question bank is used, which is developed to be answered through queries to the Freebase KB. As the main result of this work, three different Ensembles are generated, called Simple Ensemble, Ensemble with Countermeasure and Ensemble with Total Countermeasure, which are state-of-the-art for the KB-QA task considering the WebQuestionsSP database with F1-scores of 75.40%, 78.72% and 81.43% respectively.