AraVADEmo: Arabic Emotion Corpus (Ekman + VAD)

Description

AraVADEmo is a corpus of Arabic tweets annotated for discrete emotion categories (based on Ekman's basic emotions) and Valence–Arousal–Dominance (VAD) dimensions. The corpus was built by collecting Arabic-language tweets via the X API using emotion-indicative seed phrases, then manually annotating each tweet with a final emotion label and VAD ratings. Please note that we only release the tweet ID along with annotations, in compliance with X's Terms of Service, which restrict distribution of tweet content. Dataset Summary: - Total annotated tweets: 8,081 - Language: Arabic (mixed dialects; collection restricted to lang:ar) - Collection window: October – December, 2023 - Collection method: X API v2 (search_recent_tweets via Tweepy), keyword/seed-phrase search - Annotation scheme: 7-class discrete emotion label + 3-dimensional VAD categorical label: low / medium / high, positive / neutral / negative) The file includes the following columns: - tweet_id: The tweet's unique X identifier - class: Final annotated emotion label (1: Happiness, 2: Anger, 3: Surprise, 4: Fear, 5: Sadness, 6: Disgust, 7: Neutral) - Valence: positive / neutral / negative - Arousal: low / medium / high - Dominance: low / medium / high

Institutions

• Umm al-Qura University

  Mecca Region, Makkah

Categories

Funders

• Umm Al-Qura University

  Grant ID: 26UQU4340196GSSR02

Licence