contextual-vad / feature_schema.json
somukandula's picture
Train contextual VAD turn-event model
d9da30a verified
{
"feature_columns": [
"vad_prob",
"speech_ms",
"silence_ms",
"energy",
"stt_confidence",
"stable_chars",
"partial_chars",
"stable_word_count",
"words_since_pause",
"tts_playback_ms",
"tts_echo_risk",
"time_since_user_started_ms",
"time_since_assistant_started_ms",
"recent_endpoint_candidate",
"vad_active",
"ends_with_punctuation",
"ends_with_continuation",
"required_slots_filled",
"assistant_speaking",
"backchannel_like",
"expected_answer_type"
],
"numeric_features": [
"vad_prob",
"speech_ms",
"silence_ms",
"energy",
"stt_confidence",
"stable_chars",
"partial_chars",
"stable_word_count",
"words_since_pause",
"tts_playback_ms",
"tts_echo_risk",
"time_since_user_started_ms",
"time_since_assistant_started_ms",
"recent_endpoint_candidate",
"vad_active",
"ends_with_punctuation",
"ends_with_continuation",
"required_slots_filled",
"assistant_speaking",
"backchannel_like"
],
"categorical_features": [
"expected_answer_type"
],
"default_features": {
"vad_prob": 0.0,
"speech_ms": 0.0,
"silence_ms": 0.0,
"energy": 0.0,
"stt_confidence": 0.0,
"stable_chars": 0.0,
"partial_chars": 0.0,
"stable_word_count": 0.0,
"words_since_pause": 0.0,
"tts_playback_ms": 0.0,
"tts_echo_risk": 0.0,
"time_since_user_started_ms": 0.0,
"time_since_assistant_started_ms": 0.0,
"recent_endpoint_candidate": 0,
"vad_active": 0,
"ends_with_punctuation": 0,
"ends_with_continuation": 0,
"required_slots_filled": 0,
"assistant_speaking": 0,
"backchannel_like": 0,
"expected_answer_type": "open_ended"
},
"event_labels": [
"listening",
"speech_started",
"endpoint_candidate",
"turn_committed",
"user_resumed",
"interruption_started",
"interruption_confirmed",
"backchannel_detected",
"false_alarm"
]
}