About Project Taarruf
A comprehensive initiative to detect and combat digital Islamophobia
Project Overview
Project Taarruf is a 12-month initiative that aims to detect digital Islamophobia using digital tools and to raise public awareness via open-access tools, thereby enabling the development of effective policies to combat Islamophobia.
Understanding the Problem
Islamophobia is a social phenomenon that can be defined as unjustified hostility directed against Islam. Since the 1990s, much of the discrimination, attacks, and social exclusion faced by Muslims around the world have been rooted in Islamophobic thinking.
The advent of the internet and the widespread use of social media platforms have given Islamophobia a digital dimension. Digital Islamophobia is largely associated with the far right and runs parallel to hate crimes in the offline world.
Limited research on digital Islamophobia suffers from several constraints:
- Lack of interdisciplinarity
- Limited dissemination of findings
- Exclusion of media-related elements
- Neglect of social networks among Islamophobic actors
- Absence of constructive alternatives
Project Activities
Building on existing guidelines developed in relation to Islamophobia, Taarruf Project encompasses a series of activities:
Development of a specialized language model and an online chatbot operating on this model for detecting and responding to Islamophobic content.
Continuous monitoring, datafication, and visualization of Islamophobic content on X (Twitter) to track trends and patterns.
Fine-tuning the chatbot using data obtained from X monitoring to improve accuracy and effectiveness.
Continuous improvement cycle integrating insights from the web application and chatbot usage.
Outputs & Outcomes
Project Taarruf's outputs and outcomes aim to address societal challenges in two complementary ways:
Provide policymakers with data on growing societal problems through:
- Monitoring of Islamophobic content on X
- Resulting datasets and analyses
- Visualization tools for understanding trends
Contribute to addressing societal challenges by:
- Enabling correction of Islamophobic language
- Transforming harmful narratives
- Making AI-powered tools publicly available
Platform Features
View real-time data from X, including number of Islamophobic tweets, replies, and engagement metrics with customizable date ranges.
Analyze trends over time with interactive charts, identify patterns, and track spikes in Islamophobic discourse.
Automated detection and classification of Islamophobic content using specialized language models trained on relevant datasets.
Open-access tool that helps users understand and correct Islamophobic language and narratives.
Access datasets and analysis for research, policy development, and advocacy purposes.
Methodology & Approach
Our approach combines computational analysis with domain expertise:
Technical Components
- X API for data collection
- Specialized language models
- Fine-tuning on domain-specific datasets
- Interactive chatbot interface
Research-Based
- Building on existing Islamophobia guidelines
- Interdisciplinary perspective
- Evidence-based approach
- Constructive solutions
Taarruf Safety Taxonomy & Risk Guidelines
The Taarruf Project Guard Safety Taxonomy provides a comprehensive framework for identifying and classifying risks related to Islamophobic content. This taxonomy serves two critical functions:
1. Risk Classification: This document provides the taxonomy of risks that are of interest to the project – these become the classes of a classifier.
2. Risk Guidelines: These guidelines determine where the line is drawn between encouraged and discouraged outputs for each risk category in the taxonomy.
Encompasses statements that encourage or could help people plan or engage in violence against Muslims. Similarly, statements that advocate discrimination, contain slurs, or voice hateful sentiments against Muslims based on their characteristics or beliefs would also be considered inappropriate under this category.
Encompasses statements encouraging individuals (who could be underage) to engage in specific sex acts against Muslims. Similarly, sexually explicit (i.e., erotic) statements would also be considered inappropriate under this category.
Encompasses statements that specifically encourage, condone, or could help people plan or execute specific crimes involving the illegal acquisition, creation, or use of guns or illegal weapons (such as explosives, biological agents, or chemical weapons), including instructions on how to create or use them against Muslims.
Encompasses statements that specifically associate Muslims with illegally producing, transferring, or consuming regulated or controlled substances like illegal drugs, tobacco, alcohol, or cannabis.
Encompasses miscellaneous statements that encourage, condone, or could help people plan or execute specific criminal activities, like arson, kidnapping, or theft against Muslims. Items should only be considered harmful under this category when they could not be seen as violating any of the other harm types.
Encompasses statements or analytical approaches that systematically dismiss, invalidate, or exclude Islamic epistemologies, normative frameworks, or sources of knowledge as illegitimate, irrational, or irrelevant in the production of knowledge about Muslims. Similarly, approaches that reject Muslim self-understandings or ethical claims a priori—rather than through reasoned critique—would also be considered inappropriate under this category.
Encompasses statements or analytical approaches that reduce Islam solely to a fixed historical, cultural, or socio-political artifact, thereby denying its ongoing normative, ethical, or transcendental claims for Muslims. Similarly, framing Islamic beliefs or practices as outdated, obsolete, or only intelligible within a past historical context—without acknowledging their contemporary lived significance—would fall under this category.
Encompasses statements or narratives that characterize Muslims, Islamic beliefs, or Muslim social and political actors as inherently backward, anti-modern, or resistant to progress. Similarly, discursive practices that homogenize Muslims or attribute reactionary tendencies to Islam as a civilizational essence—rather than situating specific views within particular contexts—would also be considered inappropriate under this category.
Who Benefits
Access comprehensive data and evidence to inform effective policies against digital Islamophobia.
Study digital Islamophobia with interdisciplinary approaches using our datasets and tools.
Use data-driven insights to inform campaigns and efforts against Islamophobia.
Access open-source tools to understand, correct, and counter Islamophobic language.
Better understand the landscape of online hostility and support advocacy efforts.
Leverage insights for improving content moderation and safety measures.
Explore the Data
Dive into our real-time monitoring dashboard to see analysis of Islamophobic content on X and understand the trends we're tracking.