config_rubric.yaml

# General UI Configuration
title: "TTS Rubric — Dynamic Evaluation"

instructions_markdown: |
  **Welcome annotator!**
  
  Instructions for multiple aspect test
  
  Please follow these steps carefully:
  1.  Enter your unique **Annotator ID** before you begin.
  2.  Listen to each audio clip from start to finish. Please use headphones.
  3.  Rate the clip using the sliders provided based on the scoring guide.
  4.  Provide any extra details in the comments box.
  5.  Click 'Save & Next' to submit your rating and load the next clip.

# The directory where your audio files are stored.
samples_directory: "sample-audios"

# Set to 'true' to shuffle the audio files, 'false' for alphabetical order.
randomize_samples: true
# Define the evaluation criteria. The UI will be built from this list.
criteria:
  - label: "Clarity & Intelligibility"
    min: 1
    max: 5
    step: 1
    default: 3
    explanations:
      1: "Unacceptable."
      2: "Often unclear or distorted; difficult to follow."
      3: "Understandable but requires effort; some words unclear."
      4: "Mostly clear, minor issues (with fast/slow playback)."
      5: "Speech is clear, easy to understand (at all speeds)."

  - label: "Accent & Pronunciation"
    min: 1
    max: 5
    step: 1
    default: 3
    explanations:
      1: "Severe pronunciation problems; largely unintelligible."
      2: "Frequent pronunciation issues that impede understanding."
      3: "Some mispronunciations that require effort to interpret."
      4: "Minor pronunciation quirks but overall fine."
      5: "Pronunciation is natural and appropriate for the target dialect."

  - label: "Tone & Suitability"
    min: 1
    max: 5
    step: 1
    default: 3
    explanations:
      1: "Tone is inappropriate or harmful for the content."
      2: "Tone often feels off or distracting from the content."
      3: "Tone is acceptable but occasionally inappropriate."
      4: "Generally appropriate tone with small mismatches."
      5: "Tone fits the content and use-case perfectly."

  - label: "Voice quality"
    min: 1
    max: 5
    step: 1
    default: 3
    explanations:
      1: "Unusable voice quality."
      2: "Poor quality with frequent artifacts."
      3: "Noticeable quality issues but still usable."
      4: "Minor artifacts but overall high quality."
      5: "Natural, pleasant voice with no artifacts."

  - label: "Customization & Flexibility"
    min: 1
    max: 5
    step: 1
    default: 3
    explanations:
      1: "No useful customization; inflexible."
      2: "Very limited or brittle customization options."
      3: "Limited customization; acceptable for simple use-cases."
      4: "Some customization available; works well for most cases."
      5: "Highly flexible and customizable for different styles."

  - label: "Listening comfort"
    min: 1
    max: 5
    step: 1
    default: 3
    explanations:
      1: "Uncomfortable or painful to listen to."
      2: "Often fatiguing or distracting to listen to."
      3: "Some listening fatigue; tolerable for short durations."
      4: "Mostly comfortable with occasional sharpness or fatigue."
      5: "Comfortable to listen to for extended periods."