FARIAS, E. S.; http://lattes.cnpq.br/4362443431160649; FARIAS, Evelyn de Souza.
Abstract:
The sentiment analysis and opinion mining in text data extracted from online social media services has gained enough strength in recent years, making it an area very researched and disseminated worldwide. However, these processes are language dependent and the english language is the most studied one, covering a larger amount of techniques and solutions in the field. In this context, obtaining overall results in these research areas becomes quite time consuming, so with the advancement of automatic text translators and that data automatic translated to english is used by some authors, it is necessary to analyze the impact these automatic translations cost in a text classification process. The experiments performed in our study showed satisfactory results in evaluation metrics and compared to the results obtained in works by other authors in similar experiments using automatic translations of text data and sentiment polarity classification. The machine translation systems used in our study showed a trend of equally efficient translations, showing that these systems have evolved considerably in recent years. As for the automatically translated text data classification we can say that from the results obtained, automatic text translation can present good results in some scenarios and case studies. However, there is a need for experimentation with more comprehensive training data volumes in the two languages studied in this document.