DISPLACE 2024 - Speaker and Language Diarization

# DISPLACE 2024 - Speaker and Language Diarization
# Based on Kundu et al., INTERSPEECH 2024
# Paper: https://www.isca-archive.org/interspeech_2024/kundu24_interspeech.html
# Dataset: https://displace2024.github.io/
#
# Task: Speaker and language diarization in multilingual conversational audio.
# Mark speaker turn boundaries, identify speakers, and label language per segment.
#
# Guidelines:
# - Listen to the full conversation to identify distinct speakers
# - Mark temporal boundaries where speaker changes occur
# - Assign consistent speaker labels throughout the conversation
# - Identify the language spoken in each segment
# - Note any overlapping speech or code-switching

annotation_task_name: "DISPLACE 2024: Speaker and Language Diarization"
task_dir: "."

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

output_annotation_dir: "annotation_output/"
output_annotation_format: "json"

annotation_schemes:
  - annotation_type: span
    name: speaker_segments
    description: "Mark temporal speaker segments with start and end times (in seconds)"
    span_mode: temporal
    labels:
      - name: "Speaker 1"
        color: "#3B82F6"
        tooltip: "First identified speaker"
      - name: "Speaker 2"
        color: "#EF4444"
        tooltip: "Second identified speaker"
      - name: "Speaker 3"
        color: "#10B981"
        tooltip: "Third identified speaker"
      - name: "Speaker 4"
        color: "#F59E0B"
        tooltip: "Fourth identified speaker"
      - name: "Speaker 5"
        color: "#8B5CF6"
        tooltip: "Fifth identified speaker"
      - name: "Overlap"
        color: "#6B7280"
        tooltip: "Multiple speakers talking simultaneously"

  - annotation_type: radio
    name: speaker_identity
    description: "Identify the current speaker for the selected segment"
    labels:
      - name: "Speaker 1"
        tooltip: "First/primary speaker in the conversation"
        key_value: "1"
      - name: "Speaker 2"
        tooltip: "Second speaker in the conversation"
        key_value: "2"
      - name: "Speaker 3"
        tooltip: "Third speaker (if present)"
        key_value: "3"
      - name: "Speaker 4"
        tooltip: "Fourth speaker (if present)"
        key_value: "4"
      - name: "Speaker 5"
        tooltip: "Fifth speaker (if present)"
        key_value: "5"
      - name: "Unknown"
        tooltip: "Cannot determine the speaker"
        key_value: "0"

  - annotation_type: radio
    name: language
    description: "What language is spoken in this segment?"
    labels:
      - name: "Hindi"
        tooltip: "Speaker is using Hindi"
        key_value: "h"
      - name: "English"
        tooltip: "Speaker is using English"
        key_value: "e"
      - name: "Tamil"
        tooltip: "Speaker is using Tamil"
        key_value: "t"
      - name: "Telugu"
        tooltip: "Speaker is using Telugu"
        key_value: "l"
      - name: "Kannada"
        tooltip: "Speaker is using Kannada"
        key_value: "k"
      - name: "Code-switched"
        tooltip: "Speaker switches between languages within this segment"
        key_value: "c"
      - name: "Other"
        tooltip: "Language not listed above"
        key_value: "o"

  - annotation_type: radio
    name: audio_quality
    description: "Rate the audio quality for this segment"
    labels:
      - name: "Clear"
        tooltip: "Audio is clear and easy to understand"
      - name: "Acceptable"
        tooltip: "Some noise but speech is understandable"
      - name: "Poor"
        tooltip: "Significant noise or distortion"
      - name: "Unintelligible"
        tooltip: "Cannot understand the speech"

audio_display:
  show_waveform: true
  playback_controls: true
  allow_speed_control: true
  show_spectrogram: true

allow_all_users: true
instances_per_annotator: 30
annotation_per_instance: 2
allow_skip: true
skip_reason_required: false