Utilização de modelagem de tópicos para identificar os assuntos mais discutidos sobre o Haiti no twitter

Abstract

With the growth of internet access by the world’s population, large volumes of data, mainly in text format, are being shared daily. Whether through social networks, blogs or forums, this data (or collection of documents) is continuously circulated on the web by users. Twitter is one of the most used means by Internet users to disseminate information through posts. These posts, called tweets, can be analyzed, interpreted, understood or classified into groups. Because they are shared in huge numbers, it was necessary to use topic modeling techniques, a set of algorithms that help group collections of documents into topics. That said, in this work, Latent Dirichlet Allocation (LDA), as a topic modeling approach, was used on tweets discussed about Haiti in Portuguese language during January 2018 to May 2022 to discover the most recurring topics. The experiment collection was divided into 53 months for a better use of the results. The results obtained are presented through tables and graphs and are discussed monthly each year. Through these results, it is possible to see the most discussed subjects during the defined periods.