/RED-Romanian-Emotions-Dataset

A novel dataset for emotion detection from Romanian text.

Primary LanguagePythonMIT LicenseMIT

RED - Romanian Emotions Datasets

The Romanian Emotions Dataset is a Twitter based dataset annotated with emotions. Currently there are 2 releases available:

Link # of tweets # of emotions Annotation Release Date
version Data & Readme 4047 5 Single-label Sep 2021
version Data & Readme 5449 7 Multi-label Jan 2022

REDv2 is based on REDv1, which is a smaller dataset, single-labeled with 5 emotions (Anger, Fear, Joy, Sadness and Neutral). REDv2 adds Trust and Surprise, bringing the number of annotated emotions to 7, in a multi-label fashion.

The datasets are available as CSVs and/or JSONs, and are pre-split in train/dev/test. Baselines are provided in their respective folder.

If you use these datasets in your research/production, kindly cite the appropriate paper, available as bibtext.