UMUTeam at TASS 2020: Combining Linguistic Features and Machine-learning Models for Sentiment Classification

This paper describes the participation of the UMUTeam at the TASS’2020 Workshop on Sentiment Analysis, in which two tracks were proposed. The first track consists in the classification of tweets according to general sentiments of tweets written in several Spanish varieties, whereas the second task consists in a fine-grained distinction between the six basic emotions described by Ekman (2009). Our proposal is based on the usage of linguistic features alone or in combination with word-embeddings. Specifically, we test Convolutional Neural Networks and Support Vector Machines with sentence embeddings. Although our proposal did not achieve the best results, we obtained the best precision rate regarding emotion detection (Task 2) and competitive results with respect to the general sentiment classification in which tweets written in different varieties of Spanish were mixed. We consider that our proposal, despite its limitations, provides substantial benefits such as the interpretability of the results.