Emotionally-tagged video database: a systematic review

Purpose The creation of databases with emotionally labeled content is essential for research in affective computing and human-computer interaction. These databases are utilized in various contexts, including emotion detection in virtual or face-to-face classrooms, advertising, driving, gaming and entertainment environments. In this work, the emotional databases (EDBs) are analyzed in order to categorize and group them, with the aim of facilitating the search and selection of the most suitable database for future research purposes. Design/methodology/approach This paper presents a systematic literature review, identifying 70 emotional databases containing emotionally labeled videos. The content of these databases was analyzed, with a focus on their emotion models, as well as their annotation and elicitation methods. Findings The analysis revealed a predominance of categorical models and explicit annotation methods. Regarding elicitation methods, spontaneous or in-the-wild emotional databases are more developed than posed ones. Although multicultural EDBs exist, most focus on a single country or ethnicity, mainly from Eastern regions. Research limitations/implications Although five major databases were utilized and a specific set of keywords was used in the search process, this study has some limitations regarding the search strategy. On the one hand, it could be refined by incorporating synonyms, related terms and paraphrased expressions of the word "emotion" (e.g. affect, feeling, mood), which may enhance the comprehensiveness of the review. A broader set of keywords could lead to the identification of additional emotional databases that were not retrieved in the initial search, contributing to a more extensive and inclusive mapping of available resources. On the other hand, expanding the search to include additional sources such as Web of Science or Taylor & Francis could improve coverage. This improvement would further contribute to a more comprehensive and representative mapping of available resources in the field. Originality/value This article makes an original contribution by providing a review of databases that host emotionally labeled videos created between 2017 and 2022. Additionally, it proposes a set of analysis criteria for these databases, along with a series of guidelines to support researchers in selecting the most suitable emotional database for their specific research context. To facilitate the practical application of these guidelines, an interactive spreadsheet has been developed, enabling researchers to efficiently identify the most appropriate database for their needs.