Clickbait Detection (Webis Clickbait Corpus)

# Clickbait Detection
# Based on clickbait detection research and shared tasks
#
# Clickbait Definition:
# Content designed to attract attention and encourage visitors to click
# on a link, often using sensationalized, misleading, or vague headlines.
#
# Clickbait Characteristics:
# - Curiosity gap: Withholding information to create curiosity
#   ("You won't believe what happened next...")
# - Sensationalism: Exaggerated claims or emotional language
#   ("SHOCKING revelation!", "This will BLOW your mind")
# - Listicles with enticing numbers: "10 things that will change your life"
# - Forward referencing: Using "this" without context
#   ("This one trick...", "This is why...")
# - Misleading framing: Title doesn't match content
# - Emotional manipulation: Triggering outrage, fear, or excitement
#
# Non-Clickbait Characteristics:
# - Informative and specific headlines
# - Accurate representation of content
# - Professional/journalistic tone
# - Complete information provided upfront
#
# Annotation Guidelines:
# 1. Judge the headline/post in isolation (don't consider what article might say)
# 2. Focus on whether it uses manipulative techniques
# 3. A headline can be interesting without being clickbait
# 4. Sponsored content is not automatically clickbait

annotation_task_name: "Clickbait Detection"
task_dir: "."

data_files:
  - sample-data.json
item_properties:
  id_key: "id"
  text_key: "headline"

output_annotation_dir: "annotation_output/"
output_annotation_format: "json"

annotation_schemes:
  # Step 1: Binary clickbait classification
  - annotation_type: radio
    name: is_clickbait
    description: "Is this headline clickbait?"
    labels:
      - "Clickbait"
      - "Not Clickbait"
      - "Borderline"
    tooltips:
      "Clickbait": "Uses manipulative techniques to bait clicks (curiosity gap, sensationalism, misleading)"
      "Not Clickbait": "Informative, accurate, and straightforward headline"
      "Borderline": "Has some clickbait elements but not clearly manipulative"

  # Step 2: Identify clickbait techniques (if clickbait)
  - annotation_type: multiselect
    name: techniques
    description: "Which clickbait techniques are used? (Select all that apply)"
    labels:
      - "Curiosity Gap"
      - "Sensationalism"
      - "Listicle Bait"
      - "Forward Reference"
      - "Emotional Appeal"
      - "Exaggeration"
      - "Vague/Ambiguous"
      - "Question Bait"
    label_colors:
      "Curiosity Gap": "#ef4444"
      "Sensationalism": "#f97316"
      "Listicle Bait": "#eab308"
      "Forward Reference": "#22c55e"
      "Emotional Appeal": "#3b82f6"
      "Exaggeration": "#8b5cf6"
      "Vague/Ambiguous": "#06b6d4"
      "Question Bait": "#ec4899"
    tooltips:
      "Curiosity Gap": "Withholds key information to create curiosity ('You won't believe...', 'What happened next...')"
      "Sensationalism": "Uses shocking or extreme language ('EXPLOSIVE', 'DEVASTATING', 'UNBELIEVABLE')"
      "Listicle Bait": "Uses numbered lists to promise valuable content ('10 ways to...', '5 things you must know')"
      "Forward Reference": "Uses vague references like 'this', 'here's why', 'the reason' without context"
      "Emotional Appeal": "Tries to trigger strong emotions (fear, outrage, excitement) rather than inform"
      "Exaggeration": "Overstates importance or impact ('will change your life', 'everyone is talking about')"
      "Vague/Ambiguous": "Intentionally unclear to force clicking for basic information"
      "Question Bait": "Asks provocative questions designed to make readers want the answer"
    min_selections: 0
    max_selections: 8

  # Step 3: Confidence rating
  - annotation_type: likert
    name: confidence
    description: "How confident are you in your classification?"
    min_value: 1
    max_value: 5
    labels:
      1: "Not confident"
      2: "Slightly confident"
      3: "Moderately confident"
      4: "Confident"
      5: "Very confident"

allow_all_users: true
instances_per_annotator: 100
annotation_per_instance: 3
allow_skip: true
skip_reason_required: false