MARSAD: A Multi-Functional Tool for Real-Time Social Media Analysis

TL;DR

MARSAD tackles the need for Arabic-focused, real-time social media analytics by delivering a multi-functional NLP platform that processes live and archived data across sentiment, propaganda, fact-checking, and hate speech. The architecture combines a user-friendly frontend with a robust backend (MongoDB/PostgreSQL) and a Model API that orchestrates containerized transformer models (e.g., AraBERT, CamelBERT, MARERT) for multilingual analyses, including asynchronous processing and secure data handling. It supports API-key-based data scraping, archive analysis, and online search, with rich visualizations and an openly shareable annotated Arabic corpus. The work aims to democratize access to advanced Arabic NLP, enabling researchers and non-technical users to monitor and understand online discourse with real-time and retrospective insights, while addressing ethics and licensing through CC BY 4.0 and responsible-use guidelines.

Abstract

MARSAD is a multifunctional natural language processing (NLP) platform designed for real-time social media monitoring and analysis, with a particular focus on the Arabic-speaking world. It enables researchers and non-technical users alike to examine both live and archived social media content, producing detailed visualizations and reports across various dimensions, including sentiment analysis, emotion analysis, propaganda detection, fact-checking, and hate speech detection. The platform also provides secure data-scraping capabilities through API keys for accessing public social media data. MARSAD's backend architecture integrates flexible document storage with structured data management, ensuring efficient processing of large and multimodal datasets. Its user-friendly frontend supports seamless data upload and interaction.

Figures (5)

• Figure 1: System Architecture
• Figure 2: Dashboard
• Figure 3: User locations
• Figure 4: Daily Trends
• Figure 5: Post analysis