This paper evaluates automated feedback generation for ENEM Competency 5 by augmenting zero-shot prompts with an Evaluator Term Set (CTA), a term set proposed in this work and extracted from human evaluator comments on Competency 5. We compare CTA-augmented feedback against a zero-shot baseline on 46 essays with human reference feedback using BERTScore F1 and the Wilcoxon signed-rank test. CTA augmentation increases mean BERTScore F1 for both models, with statistically significant improvement in semantic similarity to human feedback for both evaluated models at α = 0.05.

This paper describes the BAMBAS team’s participation in SemEval-2024 Task 4 Subtask 1, which focused on the multilabel classification of persuasion techniques in the textual content of Internet memes. We explored a lightweight approach that does not consider the hierarchy of labels. First, we get the text embeddings leveraging the multilingual tweets-based language model, Bernice. Next, we use those embeddings to train a separate binary classifier for each label, adopting independent oversampling strategies in each model in a binary-relevance style. We tested our approach over the English dataset, exceeding the baseline by 21 percentage points, while ranking in 23th in terms of hierarchical F1 and 11st in terms of hierarchical recall.