Datasets‎ > ‎

ASTD: Arabic Sentiment Tweets Dataset


This a set of Arabic tweets containing over 10,000 entries. They are labelled as one of four classes: objective, subjective negative, subjective positive, and subjective mixed. 


  • ASTD v1.0 [600 KB] or browse and download the code and data from GitHub.
    • Includes standard splits of the data into training, validation, and testing, as well as scripts to reproduce the basic experiments described in [1].


  • This work is done jointly with Mahmoud Nabil and Amir Atiya, with data collected by the first author.


  1. Mahmoud Nabil, Mohamed Aly, and Amir Atiya. ASTD: Arabic Sentiment Tweets Dataset, EMNLP, 2015. [pdf]