News Headline Emotion Roles (GoodNewsEveryone)

# News Headline Emotion Roles (GoodNewsEveryone)
# Based on Bostan et al., LREC 2020
# Paper: https://aclanthology.org/2020.lrec-1.194/
#
# This task annotates emotions in news headlines with semantic roles:
# - Emotion: What emotion is conveyed?
# - Experiencer: Who feels the emotion?
# - Cause: What caused the emotion?
# - Target: What is the emotion directed at?
# - Cue: What words signal the emotion?
#
# Emotion Categories (Extended Plutchik):
# Joy, Sadness, Fear, Anger, Surprise (positive/negative),
# Disgust, Trust, Anticipation (positive/negative),
# Love, Pride, Guilt, Shame, Annoyance
#
# Annotation Guidelines:
# 1. First identify IF there's an emotion in the headline
# 2. Classify the emotion type
# 3. Identify semantic roles (may not all be present)
# 4. Consider both writer's emotion and reader's reaction
# 5. Headlines can evoke emotions even without explicit cues

port: 8000
server_name: localhost
task_name: "News Emotion Roles"

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

output_file: annotations.json

annotation_schemes:
  # Step 1: Emotion classification
  - annotation_type: radio
    name: emotion
    description: "What is the primary emotion in this headline?"
    labels:
      - "Joy"
      - "Sadness"
      - "Fear"
      - "Anger"
      - "Positive Surprise"
      - "Negative Surprise"
      - "Disgust"
      - "Trust"
      - "Positive Anticipation"
      - "Negative Anticipation"
      - "No emotion"
    tooltips:
      "Joy": "Happiness, delight, celebration"
      "Sadness": "Grief, sorrow, disappointment"
      "Fear": "Anxiety, worry, terror"
      "Anger": "Frustration, outrage, annoyance"
      "Positive Surprise": "Pleasant amazement, good news"
      "Negative Surprise": "Shock, disbelief at bad news"
      "Disgust": "Revulsion, disapproval"
      "Trust": "Confidence, faith, security"
      "Positive Anticipation": "Hope, excitement for future"
      "Negative Anticipation": "Dread, pessimism"
      "No emotion": "Neutral, factual headline"

  # Step 2: Emotion intensity
  - annotation_type: likert
    name: intensity
    description: "How intense is the emotion?"
    min_value: 1
    max_value: 5
    labels:
      1: "Very weak"
      2: "Weak"
      3: "Moderate"
      4: "Strong"
      5: "Very strong"

  # Step 3: Mark emotion cue
  - annotation_type: span
    name: emotion_cue
    description: "Highlight words that signal the emotion (if any)"
    labels:
      - "Emotion Cue"
    label_colors:
      "Emotion Cue": "#ef4444"
    tooltips:
      "Emotion Cue": "Words or phrases that indicate the emotion"
    allow_overlapping: false

  # Step 4: Mark cause
  - annotation_type: span
    name: cause
    description: "Highlight what CAUSED the emotion (if present)"
    labels:
      - "Cause"
    label_colors:
      "Cause": "#3b82f6"
    tooltips:
      "Cause": "The event or entity that caused the emotion"
    allow_overlapping: false

  # Step 5: Reader vs writer emotion
  - annotation_type: radio
    name: perspective
    description: "Whose emotion is this primarily?"
    labels:
      - "Writer/Subject's emotion"
      - "Intended reader reaction"
      - "Both"
      - "Unclear"
    tooltips:
      "Writer/Subject's emotion": "The emotion of people in the story"
      "Intended reader reaction": "The emotion the headline is meant to evoke in readers"
      "Both": "Both writer and reader perspective"
      "Unclear": "Cannot determine the perspective"

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