AI Annotator

The Jordan Open Source Association (JOSA) is seeking four annotators to support the enhancement of Nuha, an AI model designed to detect and classify online gender based hate speech against women, the model is specialized in the Jordanian dialect.

Nuha is an AI model developed by JOSA in various Arabic dialects; including Jordanian, Egyptian, Iraqi, and Tunisian. The model aims to support researchers and civil society organizations by identifying whether online content contains gender based hate speech and classifying its type and severity.

The Jordanian Nuha was the first model developed in 2024 as a pilot model. JOSA is currently upgrading the model into a publicly available AI tool by expanding the annotated dataset, improving annotation quality, and enhancing the model's performance to provide researchers, civil society organizations, and other stakeholders with a more accurate and reliable solution for monitoring online gender based hate speech in Jordan.

Tasks and Responsibilities

• Attend a full-day training covering technology facilitated gender based violence (TF GBV), online gender based hate speech in Jordan, social media monitoring, the annotation platform, and the annotation guidelines, process, and tools.
• Conduct a social media monitoring of specific pre-identified accounts and identify posts that include instances of online gender based hate speech.
• Annotate approximately 40,000 comments in total, with each comment independently annotated by two annotators (approximately 20,000 annotations per annotator), in accordance with the annotation guidelines.
• Incorporate revisions based on feedback from the lead annotator.
• Complete all assigned annotation tasks within the agreed timeline.

Qualifications

• Due to the nature of annotation women applicants will be prioritized.
• Fluent in Arabic language (mother tongue) and Jordanian dialect.
• Previous experience in desk research.
• Regular users of social media platforms.
• Preferably between 22 and 45 years of age, reflecting the demographic most exposed to online gender based hate speech.
• Demonstrated knowledge or experience related to technology facilitated violence against women, gender based violence, women's rights, feminism, or protection programming.
• Professional or academic background in fields such as law, gender studies, journalism, human rights, media, social sciences, or humanitarian work is preferred.
• Basic knowledge of data extraction tools and social media monitoring is a plus.
• Ability to go through huge amounts of data sets.
• Ability to work both independently and as part of a team.
• Ability to work accurately, objectively, and adhere to strict deadlines.

Notes:

• A random quality check of annotated comments will be conducted by the Lead Annotator to identify inconsistencies and minimize annotation bias.
• Annotators will receive 50 Jordanian Dinars (JOD) for every 1,000 approved annotations.
• Payments will be made in installments as shown below, or as a single payment upon completion of all assigned annotations:
  • Payment 1 (3,000 annotations): 150 JOD
  • Payment 2 (8,000 annotations): 400 JOD
  • Payment 3 (9,000 annotations): 450 JOD
• Payments are subject to 5% withholding tax, in accordance with applicable regulations.
• Amendments and corrections requested during the quality assurance process will not be counted as additional annotations.
• Annotators must submit an invoice at least 10 working days before the payment date. Payments are processed on the 15th and 25th of each month.
• Annotators are encouraged to use their own laptops for the annotation tasks.
   

How to Apply:

You can apply directly through this link: 

https://link.josa.ngo/dataannotator_en