Dataset of Arabic spam and ham tweets

Area of Science:

• Natural Language Processing (NLP)
• Data Science
• Cybersecurity
• Social Network Analysis

Background:

• The scarcity of labeled Arabic datasets poses a challenge for developing robust NLP models.
• Automatic classification of spam content is crucial for maintaining clean online communication.

Purpose of the Study:

• To introduce a novel, publicly available dataset of Arabic tweets for spam detection research.
• To provide a benchmark for evaluating and advancing Arabic NLP techniques in content classification.

Main Methods:

• Collected 132,421 Arabic tweets using the Twitter REST API (January 27, 2021 - March 10, 2021).
• Developed a Python-based crawler for data acquisition.
• Classified tweets into 'ham' (non-spam) and 'spam' categories.

Main Results:

• A comprehensive dataset of 132,421 Arabic tweets, labeled for spam/non-spam content, is now available.
• The dataset serves as a valuable resource for researchers in Arabic NLP and related fields.

Conclusions:

• This dataset addresses a critical need for Arabic NLP resources, particularly in spam detection.
• The availability of this benchmark dataset is expected to foster advancements in automated content analysis for the Arabic language.