adimyth/indic-smart-turn
Audio Classification • Updated • 82
Datasets:
audio dict | endpoint_bool bool 2
classes | text_label bool 2
classes | audio_conf float32 0.65 1 | audio_reason stringlengths 16 148 | ambiguous bool 2
classes | language stringclasses 5
values | dataset stringclasses 10
values | synthetic bool 1
class | midfiller bool 1
class | endfiller bool 1
class | spoken_text stringlengths 0 407 | speaker_id stringlengths 17 17 | session stringlengths 38 38 | chunk int32 1 206 | split stringclasses 4
values | llm_conf float32 0.3 1 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
{
"bytes": [
82,
73,
70,
70,
68,
68,
0,
0,
87,
65,
86,
69,
102,
109,
116,
32,
16,
0,
0,
0,
1,
0,
1,
0,
128,
62,
0,
0,
0,
125,
0,
0,
2,
0,
16,
0,
100,
97,
116,
... | true | true | 0.95 | Standard standalone greeting expecting a response. | false | asm | indicvoices_asm | false | false | false | হেল্ল' | S4258755000386724 | 000609d4-6ffd-4968-abb3-333db6a0c4e1_0 | 1 | train | 0.8 |
{
"bytes": [
82,
73,
70,
70,
4,
26,
1,
0,
87,
65,
86,
69,
102,
109,
116,
32,
16,
0,
0,
0,
1,
0,
1,
0,
128,
62,
0,
0,
0,
125,
0,
0,
2,
0,
16,
0,
100,
97,
116,
... | true | true | 0.88 | calling name with natural pause | false | asm | indicvoices_asm | false | false | false | ঘৰত ভাল চবৰে বিকাশ বিকাশ | S4258755000386724 | 000609d4-6ffd-4968-abb3-333db6a0c4e1_0 | 3 | train | 0.7 |
{
"bytes": [
82,
73,
70,
70,
4,
102,
0,
0,
87,
65,
86,
69,
102,
109,
116,
32,
16,
0,
0,
0,
1,
0,
1,
0,
128,
62,
0,
0,
0,
125,
0,
0,
2,
0,
16,
0,
100,
97,
116,
... | false | null | 0.85 | abrupt cutoff with rising tone | false | asm | indicvoices_asm_pausecut | false | false | false | S4258755000386724 | 000609d4-6ffd-4968-abb3-333db6a0c4e1_0 | 3 | train | 0.7 | |
{"bytes":"UklGRgQYAwBXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YeAXAwDw/+L/4P/g/9j/0f/Y/+f/6P/W/8j/z/(...TRUNCATED) | true | true | 0.88 | question asking about family members completed | false | asm | indicvoices_asm | false | false | false | বাকী ঘৰত কেনেকুৱা মা দেউতাৰ | S4258755000386724 | 000609d4-6ffd-4968-abb3-333db6a0c4e1_0 | 4 | train | 0.9 |
{"bytes":"UklGRiToAwBXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDoAwBEATcCzAG4AGAAcgFMA7AE3AT6A6wCfQ(...TRUNCATED) | true | true | 0.88 | ends on elliptical question asking about others | false | asm | indicvoices_asm | false | false | false | ভালে আছ আমি ভালেই আৰু চবৰে | S4258755000386724 | 000609d4-6ffd-4968-abb3-333db6a0c4e1_0 | 5 | train | 0.8 |
{"bytes":"UklGRiToAwBXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDoAwD///////////////////////////////(...TRUNCATED) | true | true | 0.95 | finished invitation with falling terminal intonation | false | asm | indicvoices_asm | false | false | false | "এই কি হ'ল বহুত দিন লগ কৰা নাই লগ চগ কৰিব (...TRUNCATED) | S4258755000386724 | 000609d4-6ffd-4968-abb3-333db6a0c4e1_0 | 6 | train | 0.9 |
{"bytes":"UklGRiToAwBXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDoAwCE0VLOhM+R1YTfuepE9Of6oP81BDwJyA(...TRUNCATED) | false | null | 0.85 | ends on trailing connective with rising pitch | false | asm | indicvoices_asm_pausecut | false | false | false | S4258755000386724 | 000609d4-6ffd-4968-abb3-333db6a0c4e1_0 | 6 | train | 0.9 | |
{"bytes":"UklGRiToAwBXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDoAwB8/5sAzP6oCnwiQDd8QadBfDSeF8T1B+(...TRUNCATED) | true | true | 0.95 | ends on a complete question intonation | false | asm | indicvoices_asm | false | false | false | "দে আজি আজি তই আজি আজিও আজি বিজি আছানে ন(...TRUNCATED) | S4258755000386724 | 000609d4-6ffd-4968-abb3-333db6a0c4e1_0 | 7 | train | 0.9 |
{"bytes":"UklGRkQyAwBXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YSAyAwD///////////////////////////////(...TRUNCATED) | false | null | 0.95 | Mid-thought hesitation with flat tone | false | asm | indicvoices_asm_pausecut | false | false | false | S4258755000386724 | 000609d4-6ffd-4968-abb3-333db6a0c4e1_0 | 7 | train | 0.9 | |
{"bytes":"UklGRiToAwBXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQDoAwAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA(...TRUNCATED) | true | true | 0.95 | Falling intonation and completed statement. | false | asm | indicvoices_asm | false | false | false | আজি ঘৰতে আছ | S4258755000386724 | 000609d4-6ffd-4968-abb3-333db6a0c4e1_0 | 8 | train | 0.9 |
End of preview. Expand in Data Studio
Indic Smart Turn data
Turn-boundary labels for training an Indic end-of-turn detector with the Pipecat Smart Turn v3.
Source audio is ai4bharat/IndicVoices (CC BY 4.0), conversational rows only: one speaker's side of real two-party phone calls, split into transcript segments.
Files
index/<lang>.parquet: conversational rows selected per language (shard, row, session, chunk, speaker, duration, text). No audio.labels/<lang>.jsonl: text label per segment end, from an LLM reading the ordered transcript of a session (endpoint_bool,llm_conf,label_model).labels/<lang>.audio.jsonl: Gemini audio verdict per built clip (verdict,confidence,reason), keyed bysession|chunk|dataset|md5(audio)[:10]. This is the primary label. Includes verdicts for pause-cut clips (datasetends in_pausecut).built/<lang>.parquet: 8 s 16 kHz WAV clips (audiostruct of bytes) withendpoint_bool(Gemini audio verdict, primary),text_label,audio_conf,audio_reason,ambiguous,spoken_text,speaker_id,session,chunk, and a speaker-disjointsplit(train,dev,test,test_ambiguous). Pause-cut rows (datasetends in_pausecut) are kept only when the audio verdict is incomplete. 50,421 samples in total.
Languages: Hindi, Marathi, Tamil, Kannada, Malayalam, Gujarati, Punjabi, Telugu, Assamese, Odia, Bengali (ISO-639-3 codes in file names).
How labels were made
See the project repository for the full method and code. Short version: a text LLM reads each session's transcript and labels every segment end complete/incomplete; gemini-3.7-flash then hears every built clip and gives the deciding verdict. Text and audio agree on 89.5–93.3% of segments per language; only 39–50% of pause-cut clips are truly incomplete, so pause-cuts are kept only when the audio verdict says incomplete.
Attribution: AI4Bharat (IndicVoices), SPRING Lab IIT Madras / TamilEOT, Pipecat.
- Downloads last month
- 124
Models trained or fine-tuned on adimyth/indic-smart-turn-data