Sentiment Analysis of Multilingual Texts Using Machine Learning Methods
Anton I. Kanev, Grigory A. Savchenko, Ilya A. Grishin, Denis A. Vasiliev, Emilia M. Duma
Abstract
Sentiment analysis is a topical task of evaluating the content of texts, articles and statements in order to study public opinion and the relationship between the moods of users and their phrases. At the same time, the analysis of multilingual texts is more difficult. Therefore, in this work a study was carried out with datasets in different languages, including the use of automatic translation. Various machine learning methods, several neural network architectures, and the VADER analyzer were applied. Also, NER was combined with other techniques in the work to determine the sentiment of individual entities. The authors evaluated the methods on various datasets using the F-measure. The obtained metric values show the best results for a neural network.