Polarization detection in short texts is a challenging and relevant problem in Natural Language Processing, particularly in social media environments where regional variationsand subtle discursive nuances converge. Inthis paper, we describe our participation inSubtask 1 (Spanish) of SemEval-2026 Task 9(Naseem et al., 2026a), which focuses on binary polarization classification. We evaluatetwo main strategies: lexical models based onBag-of-Words representations and regionallypre-trained Transformer models for Spanish. Inaddition, we explore a logistic stacking framework that combines lexical and contextual representations. Our experiments show that regionally adapted Transformers generally outperform purely lexical approaches, with BILMALATachieving the strongest performance in this task.The results highlight the importance of regionally aligned pre-training on social media datafor effective polarization detection in Spanish.

The information shared on social media is increasingly important; both images and text, and maybe the most popular combination of these two kinds of data are the memes. This manuscript describes our participation in Memotion task at SemEval 2020. This task is about to classify the memes in several categories related to the emotional content of them. For the proposed system construction, we used different strategies, and the best ones were based on deep neural networks and a text categorization algorithm. We obtained results analyzing the text and images separately, and also in combination. Our better performance was achieved in Task A, related to polarity classification.