Construction of Geometries Based on Automatic Text Interpretation

• Autor(a) principal: Bernardo, Miguel Sequeira de Oliveira
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/145303
• topic: Text interpretation; Word space geometry; Word2Vec; Data Science; Economy; Python; Domínio/Área Científica::Engenharia e Tecnologia::Outras Engenharias e Tecnologias

When dealing with expanding systems, like the universe or the economy, due to its constant expansion, the statistical error is remarkably high to allow an understanding of the behaviour of the system. Thus, arises the necessity to transform an expanding system into a more straightforward system to work with. In order to address the problems, a geometric word space was constructed based on automatic text interpretation. News articles and economic reports about the European Union were collected and, using Python scripts, were cleaned and used to train a Word2Vec model. The trained model created multi-dimensional word spaces from three periods of the last decades, a pre-2000 period, a 2000-2008 period and a post-2008 period. After the interpretation of the created word spaces, a difference in behaviour between the country names was noticed. All the European Union member states were getting closer to each other until 2008, but after that year there was an abrupt rupture in this trend and ever country drifted apart. This behaviour can be linked with the 2008 financial crisis, though more research is needed to confirm this behaviour and hopefully found other correlations connecting the word spaces be- haviour and the real world. An improvement in the quantity and quality of the corpus will certainly improve the accuracy of the world space and enable a better understanding of the word spaces behaviour.