MTASA

Dataset Documentation: Multi-Task Arabic Sentiment Analysis Overview This dataset contains Modern Standard Arabic (MSA) social media text (tweets and comments) annotated for three related tasks: Emotion Recognition, Sentiment Polarity, and Intention Classification. Dataset Statistics Total samples: 12644 Columns: Text, Emotion, Polarity, Intention Column Descriptions Text: The cleaned and normalized Modern Standard Arabic text from social media. Emotion: Annotated labels based on Ekman's model (happiness, sadness, anger, fear, surprise, disgust, neutral). Polarity: Sentiment labels (positive, negative, neutral). Intention: User intent labels (informative, expressive, critical/feedback, persuasive/action-oriented).