Geração de perguntas em linguagem natural a partir de bases de dados abertos e conectados: um estudo exploratório

Abstract

The accelerated growth of open and connected databases has recently been observed. There are several motivations leading to it, some bases generate data from texts automatically, other bases are built straightaway from information systems. Therefore, a numerous set of data base with a huge volume of information is now available. Thus, the possibility of its use in large scale, in systems of question and answer is observed. The question and answer systems depend on the existence of an information structure to be used as support in generating sentences and checking the answers. The current background of open and connected data provides the essential support. From literature studies, it was observed the opportunity to use more the possibilities of generating sentences in natural language from connected open databases, in different kinds of applications. In addition, several challenges have been identified to realize the effective use of this resource. So, this work aims to verify which aspects of the structure of connected open databases can be used as the support to generate questions in natural language. Since, an exploratory study was developed, and a general approach was established. Which was tested on a prototype that was able to generate natural language question sentences, supported by open and connected databases. The results were evaluated by a specialist in linguistics and were considered promising.