MASSIVE - Multilingual Intent Classification and Slot Filling

# MASSIVE - Multilingual Intent Classification and Slot Filling
# Based on FitzGerald et al., ACL 2023
# Paper: https://aclanthology.org/2023.acl-long.235/
# Dataset: https://github.com/alexa/massive
#
# MASSIVE is a 1M-example, 51-language parallel dataset for virtual-assistant
# natural language understanding (NLU), created by localizing the English
# SLURP dataset. Each spoken command is annotated with:
#   - a DOMAIN (18 scenarios) and INTENT (60 intents), and
#   - SLOTS: the spans carrying the information the assistant must act on.
#
# This showcase uses a representative subset of intents (the real task has 60)
# and a compact slot inventory. An English gloss is provided for reference.
#
# Annotation workflow:
#   1. Pick the DOMAIN (broad scenario) of the command.
#   2. Pick the INTENT (specific action requested).
#   3. Highlight each SLOT span and label what kind of value it holds.

annotation_task_name: "MASSIVE - Intent Classification & Slot Filling"
task_dir: "."

data_files:
  - sample-data.json

item_properties:
  id_key: "id"
  text_key: "utterance"

output_annotation_dir: "annotation_output/"
output_annotation_format: "json"

port: 8000
server_name: localhost

annotation_schemes:
  # Step 1: domain / scenario
  - annotation_type: radio
    name: domain
    description: "Which domain (broad scenario) does this command belong to?"
    labels:
      - "alarm"
      - "calendar"
      - "cooking"
      - "datetime"
      - "email"
      - "general"
      - "iot"
      - "lists"
      - "music"
      - "news"
      - "play"
      - "qa"
      - "recommendation"
      - "social"
      - "transport"
      - "weather"

  # Step 2: intent (representative subset of the 60 MASSIVE intents)
  - annotation_type: radio
    name: intent
    description: "What specific action is the user requesting?"
    labels:
      - "alarm_set"
      - "calendar_set"
      - "calendar_query"
      - "cooking_recipe"
      - "datetime_query"
      - "email_sendemail"
      - "email_query"
      - "iot_hue_lightchange"
      - "iot_cleaning"
      - "lists_createoradd"
      - "play_music"
      - "news_query"
      - "qa_factoid"
      - "recommendation_events"
      - "social_post"
      - "transport_query"
      - "weather_query"
      - "general_quirky"

  # Step 3: slot filling
  - annotation_type: span
    name: slots
    description: "Highlight each slot and label the type of value it holds"
    labels:
      - "time"
      - "date"
      - "location"
      - "person"
      - "device_type"
      - "media_type"
      - "app_name"
      - "food_type"
      - "weather_descriptor"
      - "number"
      - "event_name"
      - "other"
    label_colors:
      "time": "#3b82f6"
      "date": "#0ea5e9"
      "location": "#ef4444"
      "person": "#ec4899"
      "device_type": "#a855f7"
      "media_type": "#8b5cf6"
      "app_name": "#14b8a6"
      "food_type": "#f97316"
      "weather_descriptor": "#eab308"
      "number": "#22c55e"
      "event_name": "#f59e0b"
      "other": "#6b7280"

annotation_instructions: |
  You will see a short command spoken to a virtual assistant, with its
  language identifier and (for this showcase) an English gloss.
    1. Choose the DOMAIN: the broad scenario (weather, music, alarm, ...).
    2. Choose the INTENT: the specific action requested (weather_query,
       play_music, alarm_set, ...).
    3. Highlight every SLOT - the words carrying the information the assistant
       must extract - and label each with its value type (time, date,
       location, person, device_type, ...).
  Not every command has slots; some (e.g. "tell me a joke") are intent-only.

html_layout: |
  <div style="padding: 15px; max-width: 800px; margin: auto;">
    <div style="background: #ecfdf5; border: 1px solid #a7f3d0; border-radius: 8px; padding: 12px; margin-bottom: 12px;">
      <strong style="color: #065f46;">Language:</strong>
      <span style="font-size: 15px; margin-left: 8px;">{{language}}</span>
    </div>
    <div style="background: #f0f9ff; border: 1px solid #bae6fd; border-radius: 8px; padding: 16px; margin-bottom: 8px;">
      <strong style="color: #0369a1;">Command:</strong>
      <p style="font-size: 18px; line-height: 1.7; margin: 8px 0 0 0;">{{utterance}}</p>
    </div>
    <div style="color: #6b7280; font-size: 13px; font-style: italic; margin-bottom: 16px;">
      English gloss (reference only): {{gloss}}
    </div>
  </div>

allow_all_users: true
instances_per_annotator: 50
annotation_per_instance: 2
allow_skip: true
skip_reason_required: false