Nuha logo

About Us

Nuha is a tool for classifying and analyzing discourse in the Arabic language across various dialects to monitor online gender-based hate speech. Powered by AI, it is built on free, open-source technologies and was developed by the Jordan Open Source Association (JOSA) in partnership with local regional partners.

Nuha is designed to analyze online gender-based hate speech on social media through advanced algorithms. These algorithms process Arabic texts in specific dialects; such as Jordanian, Egyptian, Iraqi, and Tunisian, as well as the Kurdish language, classifying whether they contain hate speech and its subcategories, neutral discourse, or other content.

The idea behind Nuha stems from the recognition that online gender-based hate speech classification tools tailored to Arabic content are scarce. Furthermore, the content moderation tools currently used by social media platforms are weak and ineffective at monitoring hate speech in Arabic, failing to analyze and understand local and regional contexts or distinguish various forms of violence directed against women.


Is Nuha available to specific entities or individuals? And why is it important?

Nuha is accessible to all internet users, specifically:

  • Research institutions, male and female researchers.
  • Human rights and women's rights advocates.
  • Non-profit entities such as feminist and human rights organizations.
  • Policymakers.
  • male and female journalists.
  • media organizations working on gender-based violence, hate speech, and technology-facilitated gender-based violence (TFGBV).
Key Objectives of Nuha:
  • Enhance Understanding: Simplifies understanding the spread of online gender-based hate speech within textual data across various contexts.
  • Pattern Analysis: Assists in monitoring and analyzing patterns of online gender-based hate speech or bias in published content.
  • Evidence-Based Reporting: Supports building reports based on real-world data to back strategies and policy-making.
How Do I Use Nuha?

Logging In:
  1. You can log in to start using Nuha by clicking the "Start" button or the "Log In" button.
  2. Enter your email address in the designated field and click "Continue with Email." A magic link will be sent to your inbox, allowing you to sign in without creating an account or setting up a password.
  3. A confirmation message will appear: "The magic link has been sent to example@email.com! Please check your email to complete the login process."
  4. Go to your inbox and click the "Log In" button in the email you received.

  5. If you have a GitHub account, you can also sign in through it.
     

    Note: The process is safe and simple, no personal data of users is or ever will be saved!

Text Analysis:


1. Single Text Classification (One text at a time)
Enter a single Arabic text into the analysis interface, select the desired dialect, and click "Analyze."
Nuha will classify the text and determine whether it contains hate speech, along with the specific type of hate speech if present (note that types may vary by dialect due to differing regional contexts).
Additionally, Nuha provides a confidence score for its classification based on its algorithms and training sample:

  • Less than 50%: Low confidence in accuracy or correctness.
  • 60% – 70%: Somewhat low confidence.
  • 71% – 80%: Moderately confident.
  • 81% and above: Very confident.

   To analyze another text, click "Return to Analysis" and repeat the process.
Pay attention to the provided notes, as they may affect classification and analysis accuracy.

2. Bulk Text Classification (Multiple texts at once)
While analyzing a single text is useful, processing multiple texts at once can often be more efficient. You can upload a CSV file using a pre-formatted template:

  1. Go to "Upload File."
  2. Download the template to your device and select your preferred file format/extension.
  3. Copy and paste your texts into the template following the specified layout.
  4. Rename the file as desired before uploading it to the platform.

  5. Upload the file and click "Analyze."
     

    Pay attention to the requirements, primarily the maximum file size limit of 10 MB.

Analysis Results


Nuha presents text analysis results through detailed tables and analytical charts:


Detailed Tables: Display texts, classifications, platforms, confidence scores, and more. Users can filter data to select desired fields and hide non-essential information.
Analytical Charts: Available in various formats (such as tables or pie charts), giving users the freedom to specify which data types to display.
Users can download the results in all available formats to better understand hate speech patterns within their data and build data-driven reports for research or advocacy.


For a deeper understanding of Nuha and its underlying methodology, visit the Publications page to read about the methodology, as well as publications on technology-facilitated gender-based violence and online gender-based violence across different contexts in the region.


Greetings from Team Josa and all those who support Nuha, including partners, annotators, donors, and users!


Together for safe digital spaces for everyone!