This study aims to propose a solution to the problem of identifying the feeling of comments in Spanish, due to the linguistic variations existing in the different Latin American countries, expressed in social networks using as an example a political context of an Argentinian Province. To achieve this, a combination of an unsupervised machine-learning algorithm was used to do the pseudo classification, with a supervised machine-learning algorithm, for the classification model. The results show that the level of accuracy obtained is 93%, which is higher than the levels of accuracy found in previous studies. Among the contributions of the study, we can highlight the need to include a layer of pre-processing, to correct spelling errors and reduce vectorization by generating a classifier with greater precision; and a pseudo-classification process, as an alternative to manually classifying thousands of comments to achieve a dataset for training a classifier.