Размер шрифта: 

This article focuses on data triangulation as the approach to improve the accuracy of Machine Learning based sentiment analysis of media texts. Along with discussing some critical conceptual issues for machine learning-based sentiment-analysis, the paper shows the effect that data triangulation can have on the sentiment analysis of media-texts demonstrating how accuracy and precision-recall, F-scores performance metrics change when machine learning is running on triangulated vs. non-triangulated data subsets. The paper demonstrates the performance of K-nearest neighbors and Naïve Bayes algorithms on data pre-processed with ADASYN oversampling method for the task of sentiment analysis of full-text news articles. The main objectives of the research were achieved as the obtained results demonstrate the positive effect of data triangulation on the K-NN and Naïve Bayes machine learning algorithms performance. The obtained results were also confirmed for synthetically generated data during ADASYN.