AfriHate - Hate and Abusive Language for African Languages

# AfriHate - Hate and Abusive Language for African Languages
# Based on Muhammad et al., NAACL 2025
# Paper: https://aclanthology.org/2025.naacl-long.92/
# Dataset: https://github.com/AfriHate/AfriHate
#
# AfriHate is a collection of hate speech and abusive language datasets for
# 15 African languages. Native speakers familiar with the regional culture
# label each tweet with a three-way category and, for hateful tweets, the
# targeted attribute.
#
# Primary labels:
# - Hate: attacks or dehumanizes people BECAUSE of a protected group
#   attribute (ethnicity, religion, politics, gender, disability, etc.)
# - Abusive: insulting, offensive, or profane, but NOT targeting a protected
#   group (e.g. a personal insult, general vulgarity)
# - Neutral: neither hateful nor abusive (normal speech)
#
# Target attribute (only for HATE tweets):
#   ethnicity, religion, politics, gender, disability, other
#
# Annotation Guidelines:
# 1. Read the tweet in the source language (an English gloss is provided for
#    reference in this showcase only).
# 2. Choose the single best primary label. The key test for HATE is whether
#    the hostility is directed at a protected group ATTRIBUTE, not just an
#    individual.
# 3. If (and only if) you chose Hate, mark which attribute(s) are targeted.
# 4. Judge the tweet as written; do not assume missing context.

annotation_task_name: "AfriHate - African Hate and Abusive Language"
task_dir: "."

data_files:
  - sample-data.json

item_properties:
  id_key: "id"
  text_key: "text"

output_annotation_dir: "annotation_output/"
output_annotation_format: "json"

port: 8000
server_name: localhost

annotation_schemes:
  # Step 1: three-way primary label
  - annotation_type: radio
    name: moderation_label
    description: "How would you categorize this tweet?"
    labels:
      - "Hate"
      - "Abusive"
      - "Neutral"
    keyboard_shortcuts:
      "Hate": "1"
      "Abusive": "2"
      "Neutral": "3"
    tooltips:
      "Hate": "Attacks or dehumanizes people because of a protected-group attribute (ethnicity, religion, politics, gender, disability, etc.)"
      "Abusive": "Insulting, offensive, or profane, but not targeting a protected group"
      "Neutral": "Neither hateful nor abusive"

  # Step 2: targeted attribute - only fill in when the label is Hate
  - annotation_type: multiselect
    name: hate_target
    description: "If the tweet is HATE, which attribute(s) does it target? (leave blank otherwise)"
    labels:
      - "Ethnicity"
      - "Religion"
      - "Politics"
      - "Gender"
      - "Disability"
      - "Other"
    tooltips:
      "Ethnicity": "Targets an ethnic, racial, or national group"
      "Religion": "Targets a religious group or belief"
      "Politics": "Targets a political affiliation or party"
      "Gender": "Targets gender or sexual identity"
      "Disability": "Targets disability or health condition"
      "Other": "Targets some other protected attribute"

annotation_instructions: |
  You will see a tweet written in an African language, with its language
  identifier and (for this showcase only) an English gloss.

  1. Assign one primary label:
     - Hate: hostility aimed at people for belonging to a protected group
     - Abusive: offensive/insulting but not aimed at a protected group
     - Neutral: neither of the above
  2. If you chose Hate, also mark which attribute(s) it targets.

  The deciding question for Hate vs. Abusive is WHO the hostility targets:
  a protected group attribute (Hate) versus an individual or nobody in
  particular (Abusive).

html_layout: |
  <div style="padding: 15px; max-width: 800px; margin: auto;">
    <div style="background: #ecfdf5; border: 1px solid #a7f3d0; border-radius: 8px; padding: 12px; margin-bottom: 12px;">
      <strong style="color: #065f46;">Language:</strong>
      <span style="font-size: 15px; margin-left: 8px;">{{language}}</span>
    </div>
    <div style="background: #fef2f2; border: 1px solid #fecaca; border-radius: 8px; padding: 16px; margin-bottom: 8px;">
      <strong style="color: #991b1b;">Tweet:</strong>
      <p style="font-size: 17px; line-height: 1.7; margin: 8px 0 0 0;">{{text}}</p>
    </div>
    <div style="color: #6b7280; font-size: 13px; font-style: italic; margin-bottom: 16px;">
      English gloss (reference only): {{gloss}}
    </div>
  </div>

allow_all_users: true
instances_per_annotator: 50
annotation_per_instance: 3
allow_skip: true
skip_reason_required: false