aabroukh's picture
Add files using upload-large-folder tool
b136cbe verified
|
Raw History Blame Contribute Delete
5.22 kB

Firebase Data Structure for Audio Task

Document ID Format

  • Regular trials: {userId}_{sessionId}_audio_{trialId}
  • Attention check: {userId}_{sessionId}_audio_attention_check

Complete Data Structure Submitted to Firebase

1. Participant and Trial Information

  • participantId: User's Firebase Auth UID
  • promptId: {conv_id}_{turn_no} - Unique identifier for the prompt
  • trialNumber: 0 for attention check, 1-20 for regular trials
  • trialId: "attention_check" or the actual trial ID
  • isAttentionCheck: Boolean flag

2. Content Data

  • conversationHistory: Previous conversation context (if any)
  • query: The user's current query/prompt
  • audioClip1: Filename of first audio (in presentation order)
  • audioClip2: Filename of second audio (in presentation order)
  • audioLength1: Duration of first audio in seconds
  • audioLength2: Duration of second audio in seconds

3. AI/Model Ratings (Ground Truth)

  • prismRating1: PRISM score for text 1
  • prismRating2: PRISM score for text 2
  • aiRatingResponse1: AI rating for response 1
  • aiRatingResponse2: AI rating for response 2
  • aiAudioRatingResponse1: AI audio rating for response 1
  • aiAudioRatingResponse2: AI audio rating for response 2
  • aiQualityRatingResponse1: AI quality rating for audio 1
  • aiQualityRatingResponse2: AI quality rating for audio 2

4. Participant Ratings

  • participantRatingNum: Trial number + 1
  • ratingAudio1: Final response rating for audio 1 (1-100)
  • ratingAudio2: Final response rating for audio 2 (1-100)
  • qualityRatingAudio1: Final quality rating for audio 1 (1-100)
  • qualityRatingAudio2: Final quality rating for audio 2 (1-100)

5. Rating Change History (WITH AUDIO TRACKING)

  • ratingChangeHistoryAudio1: Array of objects, each containing:
    • value: Rating value (1-100)
    • timestamp: ISO timestamp when rating was released
    • audioPlaying: Boolean - was audio 1 playing when rated
    • audioCurrentTime: Current playback position in seconds
    • audioDuration: Total duration of the audio file
  • ratingChangeHistoryAudio2: Same structure for audio 2
  • qualityChangeHistoryAudio1: Same structure for quality rating
  • qualityChangeHistoryAudio2: Same structure for quality rating

6. Timing Information

  • taskStartTime: ISO timestamp when trial started
  • taskEndTime: ISO timestamp when trial was submitted
  • timestampRatingAudio1: When final rating for audio 1 was set
  • timestampRatingAudio2: When final rating for audio 2 was set
  • timestampQualityRating1: When final quality rating 1 was set
  • timestampQualityRating2: When final quality rating 2 was set
  • totalTaskDurationMs: Total time spent on trial in milliseconds

7. Order Information

  • orderPresented: Array [1,2] or [2,1] indicating presentation order
  • audioShownFirst: Which audio (1 or 2) was shown first

8. Listening Behavior Data

  • timesRelistenedAudio1: Number of times audio 1 was replayed
  • timesRelistenedAudio2: Number of times audio 2 was replayed
  • listeningTimesAudio1: Array of ISO timestamps when audio 1 was played
  • listeningTimesAudio2: Array of ISO timestamps when audio 2 was played
  • relistenTimesAudio1: Array of ISO timestamps for replays of audio 1
  • relistenTimesAudio2: Array of ISO timestamps for replays of audio 2

9. Seek/Skip Event Data

  • seekEventsAudio1: Array of seek events for audio 1, each containing:
    • timestamp: When the seek occurred
    • fromTime: Starting position before seek
    • toTime: End position after seek
    • seekDirection: 'forward' or 'backward'
    • seekDistance: Distance seeked in seconds
  • seekEventsAudio2: Same structure for audio 2
  • skipEventsAudio1: Array of skip events for audio 1, each containing:
    • timestamp: When the skip occurred
    • fromTime: Position before skip
    • toTime: Position after skip
    • skipType: 'backward' (currently only -10s implemented)
    • skipAmount: Intended skip amount (10 seconds)
    • actualSkipped: Actual amount skipped
  • skipEventsAudio2: Same structure for audio 2
  • totalSeeksAudio1: Count of seek events for audio 1
  • totalSeeksAudio2: Count of seek events for audio 2
  • totalSkipsAudio1: Count of skip events for audio 1
  • totalSkipsAudio2: Count of skip events for audio 2

10. Metadata

  • modality: Always "audio" for this task type
  • submittedAt: ISO timestamp when data was submitted to Firebase

Example Rating Change Entry with Audio Tracking

{
  "value": 75,
  "timestamp": "2025-10-02T20:30:45.123Z",
  "audioPlaying": true,
  "audioCurrentTime": 8.532,
  "audioDuration": 12.775
}

This structure captures:

  • The rating value (75/100)
  • When the rating was made
  • Whether audio was playing during rating adjustment (true)
  • Where in the audio the user was (8.5 seconds into a 12.8 second clip)

Data Validation Notes

✅ All timestamps are in ISO 8601 format ✅ All durations are in seconds (floating point) ✅ Rating values are integers from 1-100 ✅ Arrays are empty [] if no events occurred ✅ Null values only used for optional timestamp fields when events haven't occurred ✅ Audio tracking data is captured for EVERY rating/quality change event