meeting_id
string
audio_id
string
text
string
audio
audio
begin_time
float32
end_time
float32
microphone_id
string
speaker_id
string
EN2001a
AMI_EN2001a_H04_MEO069_0330297_0330718
IF YOU IF YOU S. S. H. AND THEY HAVE THIS BIG WARNING ABOUT DOING NOTHING AT ALL IN THE GATEWAY MACHINE
3,302.97
3,307.18
H04
MEO069
EN2001a
AMI_EN2001a_H00_MEE068_0414915_0415078
I'VE GOTTEN MM HARDLY ANY
4,149.15
4,150.78
H00
MEE068
EN2001a
AMI_EN2001a_H03_MEE067_0319290_0319815
IT'S YEAH I MEAN THE WAVE DATA ARE OBVIOUSLY NOT GONNA GET OFF THERE COMPLETELY
3,192.9
3,198.15
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0145515_0146152
YEAH IT'LL IT'LL PLAY THEM IN SOME ORDER IN WHICH THEY WERE SET BECAUSE OTHERWISE IT'S GONNA BE MORE ENTERTAINING
1,455.15
1,461.52
H04
MEO069
EN2001a
AMI_EN2001a_H03_MEE067_0478127_0478164
YEAH
4,781.27
4,781.64
H03
MEE067
EN2001a
AMI_EN2001a_H02_FEO065_0436920_0436957
HMM
4,369.2
4,369.57
H02
FEO065
EN2001a
AMI_EN2001a_H04_MEO069_0171941_0172087
ALL THESE FANCY PENS
1,719.41
1,720.87
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0122764_0123754
LIKE SOMETHING THAT REPRESENTS THE WHOLE SERIES IN IN A V IN A STRUCTURE VERY SIMILAR TO THE STRUCTURE IN WHICH WE REPRESENT INDIVIDUAL UM MEETINGS
1,227.64
1,237.54
H04
MEO069
EN2001a
AMI_EN2001a_H03_MEE067_0368111_0368920
'CAUSE IF WE'RE GONNA ALLOW DISJOINT SEGMENTS FOR EXAMPLE THEN HOW ARE WE GONNA KNOW WHAT'S GONNA BE IN CONTEXT AT ANY GIVEN TIME
3,681.11
3,689.2
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0292554_0293396
IS ANYONE OF YOU FOR THE FOR THE DOCUMENT FREQUENCY OVER TOTAL FREQUENCY YOU GONNA HAVE TOTAL FREQUENCIES OF WORDS THEN WITH THAT RIGHT
2,925.54
2,933.96
H04
MEO069
EN2001a
AMI_EN2001a_H03_MEE067_0296353_0296603
LIKE I DON'T KNOW COPIES OF SHAKESPEARE OR SOMETHING
2,963.53
2,966.03
H03
MEE067
EN2001a
AMI_EN2001a_H02_FEO065_0081159_0081631
I'M NOT QUITE SO WHAT IT DID YOU WANT TO DO IT I YOU JUST WANTED TO ASSIGN
811.59
816.31
H02
FEO065
EN2001a
AMI_EN2001a_H03_MEE067_0210498_0210848
AND THAT WILL OBVIOUSLY MAKE IT MUCH EASIER TO DISPLAY
2,104.98
2,108.48
H03
MEE067
EN2001a
AMI_EN2001a_H01_FEO066_0436592_0437029
UH I ORDERED ACCORDING TO THE UM STARTING TIMES OF THE UTTERANCES
4,365.92
4,370.29
H01
FEO066
EN2001a
AMI_EN2001a_H03_MEE067_0032020_0032405
'CAUSE IF WE ARE I RECKON WE SHOULD ALL READ OUR CLASSES OUT OF THE DATABASE
320.2
324.05
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0143300_0143643
AND AND PROBABLY SEPARATE TO THAT AN INFORMATION ABOUT THE DIFFERENT TOPICS LIKE THAT
1,433
1,436.43
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0043779_0044242
THAT MEANS SORT OF WE HAVE MULTIPLE LEVELS OF OF REPRESENTATION WHICH WE PROBABLY
437.79
442.42
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0276847_0276933
HMM
2,768.47
2,769.33
H04
MEO069
EN2001a
AMI_EN2001a_H01_FEO066_0189116_0189239
YEAH
1,891.16
1,892.39
H01
FEO066
EN2001a
AMI_EN2001a_H04_MEO069_0044746_0044987
THEN WE SHOULD PROBABLY FIND SOME ABSTRACTION MODEL
447.46
449.87
H04
MEO069
EN2001a
AMI_EN2001a_H03_MEE067_0031358_0031560
ARE WE STILL GONNA DUMP IT INTO A DATABASE
313.58
315.6
H03
MEE067
EN2001a
AMI_EN2001a_H03_MEE067_0405743_0405802
SAY
4,057.43
4,058.02
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0511004_0511535
LIKE WITH THE DATA STRUCTURES I'M JUST LIKE OVER THESE VAGUE IDEAS OF SOME TREES I'M F
5,110.04
5,115.35
H04
MEO069
EN2001a
AMI_EN2001a_H00_MEE068_0240098_0240136
YEAH YEAH
2,400.98
2,401.36
H00
MEE068
EN2001a
AMI_EN2001a_H04_MEO069_0194411_0194948
I THOUGHT THAT WAS THE WHOLE BEAUTY THAT LIKE YOU CAN JUST MAKE A NEW X. M. L. FILE AND SORT OF TIE THAT TO THE OTHER AND AND IT TRE
1,944.11
1,949.48
H04
MEO069
EN2001a
AMI_EN2001a_H00_MEE068_0205991_0206009
HMM
2,059.91
2,060.09
H00
MEE068
EN2001a
AMI_EN2001a_H04_MEO069_0119247_0119290
HMM
1,192.47
1,192.9
H04
MEO069
EN2001a
AMI_EN2001a_H00_MEE068_0132274_0132288
MM
1,322.74
1,322.88
H00
MEE068
EN2001a
AMI_EN2001a_H04_MEO069_0007991_0008261
THE THING IS I'M AWAY THIS WEEKEND
79.91
82.61
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0380400_0381446
SO I'M JUST WONDERING IF THERE'S WAYS TO ABANDON THE WHOLE CONCEPT OF OF MEETINGS AND SORT OF BUT JUST NOT REALLY TREATING SEPARATE MEETINGS AS TOO MUCH OF A SEPARATE ENTITY
3,804
3,814.46
H04
MEO069
EN2001a
AMI_EN2001a_H03_MEE067_0495951_0496174
SO I'D JUST BE BUILDING THE DATA STRUCTURE AGAIN
4,959.51
4,961.74
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0342264_0342453
YEAH YOU'D HAVE TO COUNT IT YOURSELF YEAH
3,422.64
3,424.53
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0304005_0304468
THEN SKIP IT BECAUSE IT'S PROBABLY SOMETHING WITH A DOT IN BETWEEN WHICH IS USUALLY NOT SOMETHING YOU WANNA HAVE AND
3,040.05
3,044.68
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0325487_0325565
THE TEMPS YEAH
3,254.87
3,255.65
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0275384_0275512
ARE THEY SPOKEN NUMBERS
2,753.84
2,755.12
H04
MEO069
EN2001a
AMI_EN2001a_H03_MEE067_0176892_0176988
WELL THAT'S EASY
1,768.92
1,769.88
H03
MEE067
EN2001a
AMI_EN2001a_H01_FEO066_0474962_0475221
THAT'S WHAT I'M GUESSING THAT'S YOU KNOW
4,749.62
4,752.21
H01
FEO066
EN2001a
AMI_EN2001a_H04_MEO069_0120444_0120583
LET'S CHECK THAT OUT
1,204.44
1,205.83
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0012177_0012665
AND THEN JUST SORT OF EVERYONE MAKE SURE EVERYONE UNDERSTAND THE INTERFACE
121.77
126.65
H04
MEO069
EN2001a
AMI_EN2001a_H01_FEO066_0484951_0484980
YEAH
4,849.51
4,849.8
H01
FEO066
EN2001a
AMI_EN2001a_H03_MEE067_0097639_0097766
IN MEMORY YEAH
976.39
977.66
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0194391_0194411
YEAH
1,943.91
1,944.11
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0212730_0213069
YEAH AND THAT'S ALSO FAIRLY EASY TO STORE ALONG WITH OUR SEGMENTS ISN'T IT
2,127.3
2,130.69
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0099586_0099696
OKAY
995.86
996.96
H04
MEO069
EN2001a
AMI_EN2001a_H01_FEO066_0441100_0441243
YEAH ONE GROUP YEAH
4,411
4,412.43
H01
FEO066
EN2001a
AMI_EN2001a_H04_MEO069_0503539_0503807
WHEN DO WE HAVE TO MEET AGAIN THEN WITH THIS
5,035.39
5,038.07
H04
MEO069
EN2001a
AMI_EN2001a_H03_MEE067_0309109_0309372
YEAH I IT WOULD BE USEFUL FOR ME AS WELL
3,091.09
3,093.72
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0000560_0000601
GOSH
5.6
6.01
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0195726_0196470
SORT OF C I WAS JUST THINKING YOU KNOW LIKE IF IF THE OVERHEAD FOR HAVING THE SAME AMOUNT OF DATA COMING FROM TWO D FILES INSTEAD OF FROM ONE FILE IS MASSIVE
1,957.26
1,964.7
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0309421_0309606
AM I THE ONLY ONE WHO NEEDS IT WITH FREQUENCIES
3,094.21
3,096.06
H04
MEO069
EN2001a
AMI_EN2001a_H01_FEO066_0333857_0334962
UM I JUST UM WONDERED SO WHO'S UH THEN DOING UM THE FREQUENCIES ON ON THE WORDS
3,338.57
3,349.62
H01
FEO066
EN2001a
AMI_EN2001a_H03_MEE067_0323471_0323505
YEAH
3,234.71
3,235.05
H03
MEE067
EN2001a
AMI_EN2001a_H00_MEE068_0463820_0464033
YEAH FOR ME IT'S BETTER IF THEY'RE BY MEETING
4,638.2
4,640.33
H00
MEE068
EN2001a
AMI_EN2001a_H04_MEO069_0025836_0026585
TH YEAH THE SEARCH IS I GUESS THE SEARCH IS SORT OF A STRANGE BEAST ANYWAY BECAUSE FOR THE SEARCH WE'RE LEAVING THE NITE X. M. L. FRAMEWORK
258.36
265.85
H04
MEO069
EN2001a
AMI_EN2001a_H01_FEO066_0466281_0466707
YEAH ONE SERIES HAS THE UM SAME THREE STARTING LETTERS
4,662.81
4,667.07
H01
FEO066
EN2001a
AMI_EN2001a_H03_MEE067_0098250_0098399
AND JUST BUILD ONE IN MEMORY
982.5
983.99
H03
MEE067
EN2001a
AMI_EN2001a_H03_MEE067_0098777_0098944
I HAVE NO IDEA
987.77
989.44
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0424718_0425031
I HAVE THAT REALLY EXCITED PIRATE COPIED THING
4,247.18
4,250.31
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0023174_0023651
SO BASICALLY APART FROM THE DISPLAY MODULE THE I THE DISPLAY ITSELF
231.74
236.51
H04
MEO069
EN2001a
AMI_EN2001a_H02_FEO065_0490168_0490205
YEAH
4,901.68
4,902.05
H02
FEO065
EN2001a
AMI_EN2001a_H01_FEO066_0162752_0162930
N UH NO NO IT'S F FOR
1,627.52
1,629.3
H01
FEO066
EN2001a
AMI_EN2001a_H04_MEO069_0087994_0088294
NO BUT I MEAN LIKE HOW HOW JASMINE DOES IT INTERNALLY I DON'T KNOW
879.94
882.94
H04
MEO069
EN2001a
AMI_EN2001a_H01_FEO066_0433402_0433692
I CAN TRY TO DO IT AND SEND IT TO YOU
4,334.02
4,336.92
H01
FEO066
EN2001a
AMI_EN2001a_H00_MEE068_0181418_0181435
RIGHT
1,814.18
1,814.35
H00
MEE068
EN2001a
AMI_EN2001a_H02_FEO065_0435874_0436021
DID YOU ALSO ORDER
4,358.74
4,360.21
H02
FEO065
EN2001a
AMI_EN2001a_H04_MEO069_0093349_0093712
SORT OF LIKE BUT THAT LIKE THE PROBLEM WITH THAT IS IT'S EASY TO DO IN THE TEXT LEVEL
933.49
937.12
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0454853_0454997
LIKE IS IT JUST THE FIRST AND THE LAST LINE
4,548.53
4,549.97
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0024471_0025049
SO THE INTERFACE IS MAINLY WHILE IT'S RUNNING JUST WORKING ON DATA THAT'S JUST LOADED FROM A FILE I GUESS
244.71
250.49
H04
MEO069
EN2001a
AMI_EN2001a_H03_MEE067_0289725_0290180
IT'S JUST LIKE BEF UNTIL THE INFORMATION DENSITY IS UP AND RUNNING
2,897.25
2,901.8
H03
MEE067
EN2001a
AMI_EN2001a_H03_MEE067_0405802_0406582
SO THEN YOU'D START WITH ALL YOUR UTTERANCES HERE AND WHEN YOU GO UP TO GET TOPIC SEGMENTS YOU GO TO HERE HERE HERE HERE HERE HERE HERE
4,058.02
4,065.82
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0372047_0372898
BECAUSE IF WE'RE DOING LIKE I THINK FOR FOR THE INFORMATION DENSITY WE UH WE SHOULD CALCULATE IT ON THE LOWEST LEVEL NOT ON THE HIGHEST
3,720.47
3,728.98
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0144018_0144100
SO
1,440.18
1,441
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0141169_0141314
AND THEY HAVE A SCORE
1,411.69
1,413.14
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0060512_0060622
MM-HMM
605.12
606.22
H04
MEO069
EN2001a
AMI_EN2001a_H02_FEO065_0271867_0272305
YES BUT WHAT ARE THE OTHER THINGS THAT'S UH SOME KIND OF NUMBER
2,718.67
2,723.05
H02
FEO065
EN2001a
AMI_EN2001a_H01_FEO066_0064517_0065545
AND THAT'S NOT SO MUCH WHAT HE MEANT WITH NOT POSSIBLY LOADING EVERYTHING WAS THAT YOU M UM LOAD ALL THE UH ANNOTATION STUFF
645.17
655.45
H01
FEO066
EN2001a
AMI_EN2001a_H03_MEE067_0070589_0070782
FOR EVERY SINGLE WORD
705.89
707.82
H03
MEE067
EN2001a
AMI_EN2001a_H00_MEE068_0149028_0149049
HMM
1,490.28
1,490.49
H00
MEE068
EN2001a
AMI_EN2001a_H04_MEO069_0217655_0217714
YEAH
2,176.55
2,177.14
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0372898_0372951
BUT LIKE 'CAUSE
3,728.98
3,729.51
H04
MEO069
EN2001a
AMI_EN2001a_H01_FEO066_0188191_0188384
FOR EXAMPLE FOR THE DIALOGUE ACTS AND SO ON
1,881.91
1,883.84
H01
FEO066
EN2001a
AMI_EN2001a_H01_FEO066_0047482_0047500
HMM
474.82
475
H01
FEO066
EN2001a
AMI_EN2001a_H03_MEE067_0138344_0139072
OKAY SO MAYBE WE SHOULD BUILD A B STORE A MEAN MEASURE FOR THE SEGMENTS AND MEETINGS AS WELL
1,383.44
1,390.72
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0139834_0140209
THEN MAYBE WE CAN MORE OR LESS USE THE SAME CODE AND JUST MAKE A FEW IFS AND STUFF
1,398.34
1,402.09
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0489772_0489847
HOW DO YOU DO THAT
4,897.72
4,898.47
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0108365_0109244
BUT I'M I'M STILL CONFUSED 'CAUSE I THOUGHT LIKE THAT'S JUST WHAT JONATHAN SAID WE DO C THAT WE CAN'T DO LIKE LOAD A MASSIVE DOCUMENT OF THAT SIZE
1,083.65
1,092.44
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0458471_0458693
OH THEN I NEED SOMETHING DIFFERENT LATER ANYWAY
4,584.71
4,586.93
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0353711_0353909
YEAH I I NEED FREQUENCY AS WELL
3,537.11
3,539.09
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0310337_0310812
WE CAN PROBABLY JUST START WITH THE JAVA HASH MAP AND LIKE JUST HASH MAP OVER IT AND SEE HOW FAR WE GET
3,103.37
3,108.12
H04
MEO069
EN2001a
AMI_EN2001a_H02_FEO065_0263972_0264145
AND THEN YEAH
2,639.72
2,641.45
H02
FEO065
EN2001a
AMI_EN2001a_H03_MEE067_0504724_0504888
DO WE HAVE TO DEMONSTRATE SOMETHING NEXT WEEK
5,047.24
5,048.88
H03
MEE067
EN2001a
AMI_EN2001a_H01_FEO066_0333444_0333587
UH TH YEAH
3,334.44
3,335.87
H01
FEO066
EN2001a
AMI_EN2001a_H03_MEE067_0412181_0412299
YEAH
4,121.81
4,122.99
H03
MEE067
EN2001a
AMI_EN2001a_H04_MEO069_0339398_0339933
I CAN PROBABLY JUST IMPLEMENT LIKE A FIVE LINE JAVA HASH TABLE FREQUENCY DICTIONARY BUILDER AND SEE
3,393.98
3,399.33
H04
MEO069
EN2001a
AMI_EN2001a_H04_MEO069_0445600_0445650
OKAY
4,456
4,456.5
H04
MEO069
EN2001a
AMI_EN2001a_H01_FEO066_0333826_0333857
'KAY
3,338.26
3,338.57
H01
FEO066
EN2001a
AMI_EN2001a_H01_FEO066_0337667_0338486
SO UM I WOULD FOR EXAMPLE NEED THE UM MOST FREQ UM FREQUENT WORDS
3,376.67
3,384.86
H01
FEO066
EN2001a
AMI_EN2001a_H03_MEE067_0232846_0232936
LIKE AFTER THIS
2,328.46
2,329.36
H03
MEE067
EN2001a
AMI_EN2001a_H00_MEE068_0026583_0026620
YEAH
265.83
266.2
H00
MEE068
EN2001a
AMI_EN2001a_H04_MEO069_0053867_0054209
AND JUST HAVE DIFFERENT LIKE FINE GRAINEDNESS LEVELS SORT OF
538.67
542.09
H04
MEO069

Dataset Card for AMI

Dataset Description

The AMI Meeting Corpus consists of 100 hours of meeting recordings. The recordings use a range of signals synchronized to a common timeline. These include close-talking and far-field microphones, individual and room-view video cameras, and output from a slide projector and an electronic whiteboard. During the meetings, the participants also have unsynchronized pens available to them that record what is written. The meetings were recorded in English using three different rooms with different acoustic properties, and include mostly non-native speakers.

Note: This dataset corresponds to the data-processing of KALDI's AMI S5 recipe. This means text is normalized and the audio data is chunked according to the scripts above! To make the user experience as simply as possible, we provide the already chunked data to the user here so that the following can be done:

Example Usage

from datasets import load_dataset
ds = load_dataset("edinburghcstr/ami", "ihm")

print(ds)

gives:

DatasetDict({
    train: Dataset({
        features: ['meeting_id', 'audio_id', 'text', 'audio', 'begin_time', 'end_time', 'microphone_id', 'speaker_id'],
        num_rows: 108502
    })
    validation: Dataset({
        features: ['meeting_id', 'audio_id', 'text', 'audio', 'begin_time', 'end_time', 'microphone_id', 'speaker_id'],
        num_rows: 13098
    })
    test: Dataset({
        features: ['meeting_id', 'audio_id', 'text', 'audio', 'begin_time', 'end_time', 'microphone_id', 'speaker_id'],
        num_rows: 12643
    })
})
ds["train"][0]

automatically loads the audio into memory:

{'meeting_id': 'EN2001a',
 'audio_id': 'AMI_EN2001a_H00_MEE068_0000557_0000594',
 'text': 'OKAY',
 'audio': {'path': '/cache/dir/path/downloads/extracted/2d75d5b3e8a91f44692e2973f08b4cac53698f92c2567bd43b41d19c313a5280/EN2001a/train_ami_en2001a_h00_mee068_0000557_0000594.wav',
  'array': array([0.        , 0.        , 0.        , ..., 0.00033569, 0.00030518,
         0.00030518], dtype=float32),
  'sampling_rate': 16000},
 'begin_time': 5.570000171661377,
 'end_time': 5.940000057220459,
 'microphone_id': 'H00',
 'speaker_id': 'MEE068'}

The dataset was tested for correctness by fine-tuning a Wav2Vec2-Large model on it, more explicitly the wav2vec2-large-lv60 checkpoint.

As can be seen in this experiments, training the model for less than 2 epochs gives

Result (WER):

"dev" "eval"
25.27 25.21

as can be seen here.

The results are in-line with results of published papers:

You can run run.sh to reproduce the result.

Supported Tasks and Leaderboards

Languages

Dataset Structure

Data Instances

Data Fields

Data Splits

Transcribed Subsets Size

Dataset Creation

Curation Rationale

Source Data

Initial Data Collection and Normalization

Who are the source language producers?

Annotations

Annotation process

Who are the annotators?

Personal and Sensitive Information

Considerations for Using the Data

Social Impact of Dataset

[More Information Needed]

Discussion of Biases

Other Known Limitations

Additional Information

Dataset Curators

Licensing Information

Citation Information

Contributions

Thanks to @sanchit-gandhi, @patrickvonplaten, and @polinaeterna for adding this dataset.

Terms of Usage

Downloads last month
2,206
Edit dataset card

Models trained or fine-tuned on edinburghcstr/ami

Space using edinburghcstr/ami 1