Dataset Viewer
Auto-converted to Parquet Duplicate
utterance_id
string
transcript
string
confidence
float64
review_status
string
agreement_score
float64
rover_confidence
float64
advisory_disagree
float64
n_models_agree
int64
callsign_matched
bool
callsign_tail
string
kagc/01-01-22/11/0
five xray pop any left taxi park this frequency
0.640278
auto
0.566667
0.854167
0.5
2
false
N125XP
kagc/01-01-22/11/1
turning left and we will be sorry that go on the runway good bye bye now bye bye
0.224074
auto
0.088889
0.527778
0.944444
1
false
N125XP
kagc/01-01-22/11/2
and five x xray pop so say again
0.356922
auto
0.306878
0.638889
0.875
1
false
N125XP
kagc/01-01-22/11/3
we are coming right next to the dollar over thnovember one two five xray papa
0.437464
auto
0.295726
0.683333
0.666667
1
true
N125XP
kagc/01-01-22/12/0
luxor six two six tango mike allegheny tower
0.56994
auto
0.303571
0.65625
0.25
1
true
N626TM
kagc/01-01-22/12/1
luxor six tango mike tower
0.402778
auto
0
0.541667
0.333333
1
false
N626TM
kagc/01-01-22/12/2
theyre like one two
0.479167
auto
0.5
0.6875
0.75
1
false
N626TM
kagc/01-01-22/12/3
november six two six tango mike
0.608466
auto
0.539683
0.714286
0.428571
2
true
N626TM
kagc/01-01-22/12/4
ahhh i l lost ten
0.5
auto
0.6
0.7
0.8
2
false
N626TM
kagc/01-01-22/12/5
november six two six tango mike allegheny tower i am showing you on the i l s for runway two eight just want to confirm your i l s two two eight
0.739454
auto
0.747126
0.858333
0.387097
2
true
N626TM
kagc/01-01-22/12/6
oh yeah well you know its the first day of the year
0.916667
auto
1
1
0.25
3
false
N626TM
kagc/01-01-22/12/7
six tango mike roger thank you sir just want to make sure were on the same page here were on the two eight cleared to land win two one zero four
0.717867
auto
0.648056
0.828125
0.322581
1
false
N626TM
kagc/01-01-22/12/8
thank you cleared to land two eight six ten mike
0.718519
auto
0.655556
0.8
0.3
1
false
N626TM
kagc/01-01-22/12/9
six tango mic bases reported at four zero zero about five minutes ago
0.779915
auto
0.820513
0.903846
0.384615
2
false
N626TM
kagc/01-01-22/12/10
all alright six thing bye
0.216667
auto
0.15
0.5
1
1
false
N626TM
kagc/01-01-22/12/11
nice six tango mic stay parking
0.50463
auto
0.388889
0.625
0.5
1
false
N626TM
kagc/01-01-22/12/12
back to
0.344444
auto
0.333333
0.5
0.8
2
false
N626TM
kagc/01-01-22/12/13
one player
0.166667
auto
0
0.5
1
1
false
N626TM
kagc/01-01-22/12/14
six tango mike taxi to corporate air via golf left on golf and then left on alfa its just off the west side of the tower
0.787987
auto
0.675214
0.842593
0.153846
1
false
N626TM
kagc/01-01-22/12/15
hold down
0.229167
auto
0
0.6875
1
1
false
N626TM
kagc/01-01-22/13/0
morning car love quest six three two one i l s two eight
0.606838
auto
0.685897
0.826923
0.692308
1
false
LBQ651
kagc/01-01-22/13/1
with six one allegheny tower good morning wind one seven zero at five runway two eight clear to land
0.819444
auto
0.783333
0.875
0.2
2
false
LBQ651
kagc/01-01-22/13/2
two eight acres of land left quest six fifty one
0.550617
auto
0.451852
0.7
0.5
1
false
LBQ651
kagc/01-01-22/13/3
sixety one just be advised just in the last couple of minutes weve had a cloud move across the approach of runway two eight it extends roughly starts roughly at the numbers of two eight and then east im not sure how far east where exactly youre gonna enter that r v r reads three zero zero zero at touchdown
0.89786
auto
0.928962
0.967742
0.203125
3
false
LBQ651
kagc/01-01-22/13/4
ive got six fifty one roger
0.759259
auto
0.777778
0.833333
0.333333
2
false
LBQ651
kagc/01-01-22/13/5
do you want the taxi to departure via golf alfa this frequency
0.590598
auto
0.5
0.733333
0.461538
1
false
LBQ651
kagc/01-01-22/13/6
also offer perfect youre welcome youre welcome
0.18254
auto
0.047619
0.5
1
1
false
LBQ651
kagc/01-01-22/13/7
air sixty one do you have any powers from final
0.706818
auto
0.7
0.875
0.454545
2
false
LBQ651
kagc/01-01-22/13/8
one zero brook air time
0.166667
auto
0
0.5
1
1
false
LBQ651
kagc/01-01-22/13/9
visible right now
0.388889
auto
0.5
0.666667
1
1
false
LBQ651
kagc/01-01-22/13/10
roger that how far out to the east does that cloud extend roughly what it where did you enter the cloud
0.792797
auto
0.838889
0.920455
0.380952
3
false
LBQ651
kagc/01-01-22/13/11
about
0.277778
auto
0.333333
0.5
1
2
false
LBQ651
kagc/01-01-22/13/12
two and a half miles out
1
auto
1
1
0
3
false
LBQ651
kagc/01-01-22/13/13
thank you very much and
0.866667
auto
1
1
0.4
3
false
LBQ651
kagc/01-01-22/13/14
ground does have an alfa on zero decimal seven whenever you want it
0.694444
auto
0.717949
0.826923
0.461538
2
false
LBQ651
kagc/01-01-22/13/15
aircraft one seven bye
0.180556
auto
0
0.541667
1
1
false
LBQ651
kagc/01-01-22/14/0
its for departure see ya
0.613889
auto
0.766667
0.875
0.8
2
false
LBQ651
kagc/01-01-22/14/1
far departure left at six fifty one good day
0.481481
auto
0.333333
0.666667
0.555556
1
false
LBQ651
kagc/01-01-22/27/0
takeoff cross seventeen i l s two eight
0.344136
auto
0.074074
0.583333
0.625
1
false
FDY307
kagc/01-01-22/27/1
one seven allegheny tower runway two eight clear to land wind one six zero five
0.70463
auto
0.634722
0.8125
0.333333
1
false
FDY307
kagc/01-01-22/27/2
quadra line two eight corona seventy one
0.437831
auto
0.349206
0.678571
0.714286
1
false
FDY307
kagc/01-01-22/27/3
chrono seventeen id appreciate any base reports and wind on final thank you
0.82677
auto
0.783883
0.910714
0.214286
2
false
FDY307
kagc/01-01-22/27/4
what watch out
0.203704
auto
0.111111
0.5
1
1
false
FDY307
kagc/01-01-22/28/0
0.277778
auto
0.333333
0.5
1
2
false
CNS17
kagc/01-01-22/28/1
runner 17th going to atlantic and the winds were pretty much due south we were showing like nineteen knots on the approach and we broke out at about six zero zero agl
0.759745
auto
0.792627
0.892857
0.40625
2
false
CNS17
kagc/01-01-22/28/2
oscar seventeen thank you for that left at alfa two straight and only for atlantic with me
0.651727
auto
0.598039
0.833333
0.47619
1
false
CNS17
kagc/01-01-22/28/3
three two one
0.231481
auto
0.111111
0.583333
1
1
false
CNS17
kagc/01-01-22/28/4
process
0.166667
auto
0
0.5
1
1
false
CNS17
kagc/01-01-22/29/0
metafix of fact six tango tap it is for departure see it
0.456019
auto
0.388889
0.645833
0.666667
1
false
N626TM
kagc/01-01-22/29/1
see you later
0.361111
auto
0.333333
0.75
1
2
false
N109HQ
kagc/01-01-22/30/0
pop driver were switching good day
0.305556
auto
0.047619
0.535714
0.666667
1
false
N263PC
kagc/01-01-22/30/1
0.166667
auto
0
0.5
1
1
false
N263PC
kagc/01-01-22/30/2
bfwatch tv two zero two one
0.222222
auto
0
0.5
0.833333
1
false
N263PC
kagc/01-01-22/34/0
luxor five xray papa contact pissar departure one nine three five gday
0.408832
auto
0.232906
0.576923
0.583333
1
false
N125XP
kagc/01-01-22/34/1
maintain thirty five contact departure medevac one two five xray papa
0.944444
auto
0.878788
0.954545
0
3
true
N125XP
kagc/01-01-22/34/2
five xray papa climb maintain three zero zero zero three zero zero zero please and call nineteen thirty one nine three five
0.584255
auto
0.45873
0.703125
0.409091
1
false
N125XP
kagc/01-01-22/34/3
climb and maintain three zero zero zero and contact departure medivac one two five xray papa
0.866319
auto
0.770833
0.890625
0.0625
2
true
N125XP
kagc/01-01-22/41/0
clipper jet nine six six assume the base is around two zero zero feet
0.930556
auto
0.952381
0.982143
0.142857
3
false
N966CB
kagc/01-01-22/41/1
yeah that thats about right
0.8
auto
1
1
0.6
3
false
N966CB
kagc/01-01-22/41/2
lufthansa nine six six rudder
0.601852
auto
0.555556
0.75
0.5
1
false
N966CB
kagc/01-01-22/41/3
contact pittsburgh departure one one nine decimal three five good morning or correction good day good day
0.803922
auto
0.705882
0.882353
0.176471
2
false
N966CB
kagc/01-01-22/41/4
nineteen thirty five good day
0.387302
auto
0.133333
0.6
0.571429
1
false
N966CB
kagc/01-01-22/41/5
nine six six
0.814815
auto
0.777778
0.916667
0.25
2
true
N966CB
kagc/01-01-22/42/0
chrono seventeen contact pittsburgh departure one one nine decimal three five good day nineteen thirty five happy new year colonel seventeen
0.605873
auto
0.380952
0.77
0.333333
1
false
NKS722
kagc/01-01-22/47/0
0.277778
auto
0.333333
0.5
1
2
false
N425ZZ
kagc/01-01-22/47/1
november four alfa go say parking atlantic
0.542328
auto
0.333333
0.722222
0.428571
1
false
N644AT
kagc/01-01-22/47/2
before alfa tango left on alfa two straight in the atlantic with me and do you have any updates for prior to performing
0.793588
auto
0.750329
0.847826
0.217391
2
false
N644AT
kagc/01-01-22/47/3
yeah air vistajl two fifty
0.411111
auto
0.233333
0.6
0.6
1
false
N425ZZ
kagc/01-01-22/47/4
it was quite a animal
0.45
auto
0.6
0.75
1
1
false
N644AT
kagc/01-01-22/47/5
on traffic
0.166667
auto
0
0.5
1
1
false
N425ZZ
kagc/01-01-22/47/6
roger that
0.425926
auto
0.111111
0.666667
0.5
1
false
RPA4886
kagc/01-01-22/47/7
0.166667
auto
0
0.5
1
1
false
N425ZZ
kagc/01-01-22/53/0
0.166667
auto
0
0.5
1
1
false
JIA5418
kagc/01-01-22/53/1
0.166667
auto
0
0.5
1
1
false
JIA5418
kagc/01-01-22/53/2
cleared thats land runway one
0.247613
auto
0.002604
0.740234
1
1
false
N331RC
kagc/01-01-22/53/3
planer one romeo park and corporate
0.568783
auto
0.492063
0.714286
0.5
1
false
N331RC
kagc/01-01-22/53/4
and one romeo charlie left turn taxi alfa four
0.736237
auto
0.627273
0.854167
0.272727
1
false
N331RC
kagc/01-01-22/53/5
one yeah
0.277778
auto
0.333333
0.5
1
2
false
JIA5418
kagc/01-01-22/56/0
pittsburgh jet seven nine elliot tower runway one zero fly runway heading clear cleared takeoff wind calm
0.741013
auto
0.678922
0.838235
0.294118
1
false
N47EK
kagc/01-01-22/56/1
this fistro sending out we need about a minute to do our check here
0.656448
auto
0.616667
0.78125
0.428571
1
false
RPA4641
kagc/01-01-22/56/2
pajette seven nine roger roger ready
0.579365
auto
0.5
0.666667
0.428571
1
false
N47EK
kagc/01-01-22/57/0
pittsburgh jet seven nine contact pittsburgh departure one one nine decimal three five good day
0.789683
auto
0.585714
0.85
0.066667
2
false
RPAA4659
kagc/01-01-22/57/1
roger nineteen thirty five vistarug jet seventy nine
0.551587
auto
0.571429
0.75
0.666667
1
false
RPAA4659
kagc/01-01-22/67/0
0.166667
auto
0
0.5
1
1
false
CNS242
kagc/01-01-22/67/1
heli academy information juliett it is all current hop to meter two nine six five
0.390741
auto
0.155556
0.616667
0.6
1
false
CNS242
kagc/01-01-22/68/0
way in sight for conos two four two roger
0.541975
auto
0.37037
0.7
0.444444
1
true
CNS242
kagc/01-01-22/69/0
0.222222
auto
0
0.5
0.833333
1
false
CNS242
kagc/01-01-22/69/1
for for two were going to atlantic
0.52381
auto
0.571429
0.75
0.75
1
false
CNS242
kagc/01-01-22/69/2
two four two you can continue down the runway turn left alfa two that thatll take you straight into atlantic
0.848413
auto
0.866667
0.916667
0.238095
3
true
CNS242
kagc/01-01-22/69/3
alright down the runway alfa two straight into atlantic and viz has gotten much better
0.686356
auto
0.620833
0.838235
0.4
2
false
CNS242
kagc/01-01-22/69/4
perfect
1
auto
1
1
0
3
false
SWA2795
kagc/01-01-22/69/5
five miles left
0.1875
auto
0
0.5625
1
1
false
LXJ96
kagc/01-01-22/69/6
affirmative
1
auto
1
1
0
3
false
SWA2795
kagc/01-01-22/69/7
gotcha
0.361111
auto
0.333333
0.75
1
2
false
CNS242
kagc/01-01-22/69/8
six now
0.351852
auto
0.222222
0.5
0.666667
1
false
LXJ96
kagc/01-01-22/69/9
i dont know its just overnight change their name
0.645062
auto
0.768519
0.833333
0.666667
2
false
LXJ96
kagc/01-01-22/69/10
alrighty
0.694444
auto
0.333333
0.75
0
2
false
CNS242
kagc/01-01-22/76/0
romeo golf are you going to link
0.730159
auto
0.619048
0.821429
0.25
2
false
null
kagc/01-01-22/76/1
well go to atlantic yes maam
0.353175
auto
0.380952
0.678571
1
1
false
null
kagc/01-01-22/76/2
atlantic roger romeo golf you can turn left alfa two straight into atlantic with me
0.946296
auto
0.888889
0.95
0
3
false
null
End of preview. Expand in Data Studio

TartanAviation ATC ASR Labels

Machine transcripts and confidence scores for twangodev/tartanaviation-atc-adsb-utterances. A 1:1 labels-only add-on (no audio): one row per source utterance, same shards and row order, keyed by utterance_id.

531,050 labels · 184 shards · 100% coverage · ensemble ASR + weighted ROVER + ADS-B callsign snap · ~326 human-reviewed. Built with readback.

Usage

Join 1:1 onto the source. Rows are aligned and in the same order:

from datasets import load_dataset, concatenate_datasets

src = load_dataset("twangodev/tartanaviation-atc-adsb-utterances", split="train")
lab = load_dataset("twangodev/tartanaviation-atc-labels", split="train")

joined = concatenate_datasets([src, lab], axis=1)   # or join on utterance_id
joined[0]["audio"], joined[0]["transcript"], joined[0]["confidence"]

Schema

column type notes
utterance_id string join key; matches the source 1:1
transcript string fused transcript; "" for non-speech; human edit when review_status=edited
confidence float quality score in 0 to 1 (ranking, not calibrated; see below)
review_status string auto / verified / edited / rejected / non_speech
agreement_score float inter-model agreement, 0 to 1
rover_confidence float ROVER mean winning-vote fraction, 0 to 1
advisory_disagree float / null disagreement of an out-of-vote model with the fused text
n_models_agree int size of the agreeing plurality
callsign_matched / callsign_tail bool / string whether an ADS-B callsign was snapped in, and which tail

review_status: auto (ensemble only, the vast majority) · verified (human accepted) · edited (human corrected) · rejected (human flagged unusable) · non_speech (silence/noise, transcript=""). Human verdicts set confidence to 1.0 (0.0 for rejected).

How labels were made

Three ASR models, parakeet-tdt-0.6b-v2, canary-qwen-2.5b, and whisper-large-v3-atco2-asr, fused with weighted ROVER, callsigns snapped from ADS-B, and ~326 clips human-reviewed. confidence is the mean of agreement_score, rover_confidence, and 1 - advisory_disagree. These are machine labels, not gold ground truth (except verified/edited).

Confidence

confidence ranks labels reliably but is not a calibrated probability. It was tuned on a worklist of deliberately hard or ambiguous review clips, not a random sample. On that (hard) slice the human accept-rate was ~91% at >=0.8 and ~97% at >=0.9, so true rates on the full corpus are higher. Pick your own cutoff, or use the raw signals and review_status to build your own filter.

License

CC-BY-4.0, inherited from CMU AirLab's TartanAviation. No audio is redistributed, only derived text labels. Labels generated by nvidia/parakeet-tdt-0.6b-v2 (CC-BY-4.0), nvidia/canary-qwen-2.5b (CC-BY-4.0), and jlvdoorn/whisper-large-v3-atco2-asr (Apache-2.0); model outputs are unrestricted. Credit CMU AirLab and cite:

@article{patrikar2024tartanaviation,
  title={TartanAviation: Image, Speech, and ADS-B Trajectory Datasets for Terminal Airspace Operations},
  author={Jay Patrikar and Joao Dantas and Brady Moon and Milad Hamidi and Sourish Ghosh and Nikhil Keetha and Ian Higgins and Atharva Chandak and Takashi Yoneyama and Sebastian Scherer},
  year={2024},
  eprint={2403.03372},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  url={https://arxiv.org/pdf/2403.03372.pdf}
}
Downloads last month
-

Models trained or fine-tuned on twangodev/tartanaviation-atc-labels

Collection including twangodev/tartanaviation-atc-labels

Paper for twangodev/tartanaviation-atc-labels