Inference of topics with latent dirichlet allocation for open government data

Abstract

Open government data can be considered as an important initiative of institutions of civil society, promoting transparency and allowing its reuse as an input in the development of innovation projects. However, it is common for certain databases to require the application of specific treatments, so that the data can be used more efficiently, such as the case of classification using Data Mining. In this scenario, this paper presents an automatic topic inference proposal using the Latent Dirichlet Allocation method to classify cultural projects in their thematic areas, by identifying the similarity in their data. The results demonstrate the feasibility of the approach in the context of open government data.