Answers are grounded in this paper and cite their sources.
Emotion recognition (ER) is a preliminary step towards endowing emotional intelligence to machines, which learn it from data. Emotion data resources are pivotal resources for studying and building ER models and an important determinant for their advancement. Brazilian Portuguese (BP) is a dialect of the 8th most spoken language in the world and the 7th globally used on the web, yet it is a low-resource language. Handful of reviews exists citing BP-ER data resources but none are based exclusively on BP nor cover all research literature. This exploratory review assesses the current status of the available BP-ER data resources for advancing ER models in BP. We extensively explored emotion studies among all research publication types up to 2024 and provide an overview of the existing ER data resources in speech, facial expression, and text modalities useful for computation purposes. Overall, 59 data resources were discovered and are listed with the details of emotions included, sources employed in their creation, and their availability under respective modality. The reported data resources are smaller in size and less than 60% are available either openly or on request. A unanimous observation in all the works that created and studied these resources highlights their scarcity. As research with adequate resources will be beneficial for advancing ER modeling, we emphasize the need for the larger in-the-wild multimodal ER data resources in BP. We believe our review will be beneficial in guiding future studies on the aspects of building corpora for advancing emotion recognition modeling in BP, as well as for any low-resource language studies.
No takes yet. Share an insight, caveat, or question.
Joshi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: