Datasets:

Modalities:
Tabular
Text
Formats:
json
Libraries:
Datasets
Dask
License:
words
stringlengths
0
4.03k
end_time
float64
0.46
2.95k
speaker
stringclasses
4 values
start_time
float64
0
2.93k
session_id
stringclasses
100 values
ummm so what do you do
32.72
speaker2
30.51
session_02d73d95
i'm a medical student at drexel hmmm
38.32
speaker1
32.26
session_02d73d95
oh at penn or temple
36.52
speaker2
34.71
session_02d73d95
okay
39.32
speaker2
37.21
session_02d73d95
appoding
41.77
speaker2
40.51
session_02d73d95
twenty five
42.47
speaker1
40.81
session_02d73d95
sorry
46.57
speaker2
45.26
session_02d73d95
ummm
48.32
speaker2
47.76
session_02d73d95
did you like medical school
51.87
speaker2
50.61
session_02d73d95
yeah i really like it
53.07
speaker1
51.26
session_02d73d95
ummm are you still in classes or
56.12
speaker2
53.76
session_02d73d95
ummm i'm doing a master's in public health this year so i'm taking a year away and i haven't started yet so i start in two weeks uhhh excited be a nice change of pace
68.67
speaker1
55.31
session_02d73d95
yeah
59.32
speaker2
58.86
session_02d73d95
are you excited and nervous
65.42
speaker2
62.36
session_02d73d95
yeah what have you been doing since
70.07
speaker2
68.01
session_02d73d95
hmmm
70.07
speaker1
69.56
session_02d73d95
uhhh i have a part time research job
73.32
speaker1
71.01
session_02d73d95
and uhhh play rugby and that's about it enjoying the summer
81.07
speaker1
74.41
session_02d73d95
yeah
78.87
speaker2
77.41
session_02d73d95
yeah
84.72
speaker1
81.76
session_02d73d95
cool
83.57
speaker2
82.16
session_02d73d95
ummm what kind of research uhhh
86.37
speaker2
84.16
session_02d73d95
it is on literacy and access to ummm pediatric care services and preschool services
96.82
speaker1
85.41
session_02d73d95
so you work with little kids
101.37
speaker2
99.06
session_02d73d95
i work with little kids but it's about uhhh getting sites in the norristown area to participate in uhhh research so that
110.57
speaker1
100.26
session_02d73d95
families with low literacy can have better access to care so it's
115.32
speaker1
111.36
session_02d73d95
more like interacting with uhhh preschool directors and pediatricians
121.57
speaker1
116.51
session_02d73d95
gotcha that sounds fun though so you're up in norristown
128.72
speaker2
120.21
session_02d73d95
yeah every once in a while up in norristown walking around to different sites
130.22
speaker1
124.66
session_02d73d95
yeah
130.22
speaker2
129.76
session_02d73d95
ummm norristown i think gained some notity this summer because
135.82
speaker2
131.01
session_02d73d95
there's a bike that goes you know from the city up to valley uhhh forge and i think
142.02
speaker2
136.56
session_02d73d95
bikers were getting jumped in north
147.57
speaker2
142.76
session_02d73d95
oh really i i bike everywhere i didn't know that
147.52
speaker1
144.46
session_02d73d95
i think i didn't read i didn't read it somebody who reads like
152.57
speaker2
148.86
session_02d73d95
message boards was telling me about that
156.07
speaker2
153.36
session_02d73d95
ummm where'd you go to college
160.62
speaker2
157.56
session_02d73d95
i went to penn undergrad
162.22
speaker1
160.16
session_02d73d95
ummm
164.07
speaker2
163.36
session_02d73d95
so you
166.17
speaker2
165.26
session_02d73d95
yeah really what year did you graduate
171.07
speaker1
165.81
session_02d73d95
so did i yeah
169.62
speaker2
167.71
session_02d73d95
uhhh oh wait
174.82
speaker2
170.31
session_02d73d95
okay was oh six so yeah stuck around two in philly
177.22
speaker1
171.76
session_02d73d95
did you
177.77
speaker2
176.06
session_02d73d95
you did pre med
180.32
speaker2
178.76
session_02d73d95
uhhh yes i did pre med bbb
183.07
speaker1
179.41
session_02d73d95
cool did you take any an anthor courses
184.77
speaker2
182.06
session_02d73d95
i didn't unfortunately i wish i had didn't though yeah
188.02
speaker1
183.71
session_02d73d95
really
187.57
speaker2
186.26
session_02d73d95
yeah i did anthro and we had that we have physical
192.82
speaker2
188.86
session_02d73d95
or biological anthropology requirements so there's always bbv kids and health and societies too
202.92
speaker2
193.76
session_02d73d95
right hang around
199.32
speaker1
197.66
session_02d73d95
well what's bbb like is it a lot of cognitive stuff
210.57
speaker2
206.01
session_02d73d95
yeah it's about uhhh it's more less cognitive science more neuroscience like biology of the brain and neuropathways and stuff like that so it's bio and with a focus on neurobio but then you have to take some psych classes in there so
225.82
speaker1
209.76
session_02d73d95
that's more of
212.27
speaker2
211.51
session_02d73d95
but what's the difference between cognitive science and neuroscience
228.82
speaker2
223.76
session_02d73d95
i think cognitive science is more about how the mind works and processes and analyzes things versus like the biology of of like neurons
244.07
speaker1
227.51
session_02d73d95
got it
243.12
speaker2
242.01
session_02d73d95
ummm
245.32
speaker2
244.76
session_02d73d95
do you like buying
247.77
speaker2
246.51
session_02d73d95
yeah i really like bio
249.47
speaker1
247.06
session_02d73d95
ummm
251.57
speaker2
251.01
session_02d73d95
i i have a friend who who ummm just finished school for for biology and she always friends a really
258.72
speaker2
252.76
session_02d73d95
ummm hard to find people that she can talk about her like academic interests because i think she feels like everybody she meets at parties and like in the humanities and stuff
271.57
speaker2
259.51
session_02d73d95
right right yeah i could i could see that it's hard to talk about science in a casual like conversational way but yeah all my friends were in humanity so
282.57
speaker1
269.96
session_02d73d95
right
280.07
speaker2
279.46
session_02d73d95
yeah
282.57
speaker2
281.96
session_02d73d95
ummm
285.07
speaker2
284.36
session_02d73d95
what uhhh rugby you play in a in a league
292.17
speaker2
286.76
session_02d73d95
rugby
290.62
speaker1
288.96
session_02d73d95
i started playing at penn ummm on a club team and now i play for the philadelphia women's team so yeah that's how i yeah got here
304.12
speaker1
291.51
session_02d73d95
oh wow
297.37
speaker2
296.76
session_02d73d95
oh with ummm
300.07
speaker2
298.71
session_02d73d95
okay ummm oh man i think i met i think a lisa ummm one of the other people back there was excited about having some rugby players come to me so they could steal eat from her ummm
317.52
speaker2
302.31
session_02d73d95
yeah
316.17
speaker1
315.51
session_02d73d95
what else are you into do you like music
321.57
speaker2
319.26
session_02d73d95
i do i do
322.82
speaker1
321.16
session_02d73d95
ummm do you go to a lot of shows
324.92
speaker2
323.31
session_02d73d95
ummm not a couple a year but i started to go more recently more frequently more recently i went to a show at the tritone on friday it was uhhh the u pes really cool do you know that band
343.77
speaker1
325.61
session_02d73d95
no that's a good name
342.62
speaker2
341.46
session_02d73d95
i think they're a local band
346.82
speaker1
345.61
session_02d73d95
but
349.02
speaker1
348.61
session_02d73d95
ummm seen any good movies recently
351.12
speaker2
348.61
session_02d73d95
i saw uhhh adventureland last night uhhh it's this like kind of
358.97
speaker1
350.76
session_02d73d95
what's that
356.32
speaker2
355.06
session_02d73d95
collegi like romantic comedy sort of movie ummm
365.57
speaker1
360.31
session_02d73d95
i can't name any of the actors for you because i don't know their names but it was uhhh it was pretty good
371.82
speaker1
367.01
session_02d73d95
so five hundred days of summer
374.32
speaker1
372.51
session_02d73d95
earlier in the in the summer yeah i really liked it
377.57
speaker1
375.01
session_02d73d95
oh was that good
376.32
speaker2
375.31
session_02d73d95
ummm
381.57
speaker2
380.76
session_02d73d95
that wasn't filmed in philly was it yeah
387.82
speaker2
382.51
session_02d73d95
five hundred ace i think it was filmed in la i'm not sure i think it's set in
391.27
speaker1
384.71
session_02d73d95
oh okay 'cause i i remember it i think it debued at the philadelphia film festival yeah which i guess maybe that's why i thought it was filmed in philly but
400.02
speaker2
388.51
session_02d73d95
oh really
395.32
speaker1
394.51
session_02d73d95
i don't think so
399.87
speaker1
398.76
session_02d73d95
yeah it was pretty good soundtrack hmmm
403.82
speaker1
400.86
session_02d73d95
yeah
404.02
speaker2
400.86
session_02d73d95
so uhhh where do you where do you live now
410.72
speaker2
408.01
session_02d73d95

Dataset Name: Dataset for ASR Speaker-Tagging Corrections (Speaker Diarization)

Description

  • This dataset is pairs of erroneous ASR output and speaker tagging, which are generated from a ASR system and speaker diarization system. Each source erroneous transcription is paired with human-annotated transcription, which has correct transcription and speaker tagging.
  • SEGment-wise Long-form Speech Transcription annotation (SegLST), the file format used in the CHiME challenges

Example) session_ge1nse2c.seglst.json

[
...
    {
        "session_id": "session_ge1nse2c",
        "words": "well that is the problem we have erroneous transcript and speaker tagging we want to correct it using large language models",
        "start_time": 181.88,
        "end_time": 193.3,
        "speaker": "speaker1"
    },
    {
        "session_id": "session_ge1nse2c",
        "words": "it seems like a really interesting problem I feel that we can start with very simple methods",
        "start_time": 194.48,
        "end_time": 205.03,
        "speaker": "speaker2"
    },
...
]

Structure

Data Split

The dataset is divided into training and test splits:

  • Training Data: 222 entries
    • 2 to 4 speakers in each session
    • Approximately 10 ~ 40 mins of recordings
  • Development Data: 13 entries
    • 2 speakers in each session
    • Approximately 10 mins of recordings
  • Evaluation Data: 11 entries
    • 2 speakers in each session
    • Approximately 10 mins of recordings

Keys (items)

  • session_id: "session_ge1nse2c",
  • words: Transcription corresponding to the time stamp (start, end).
  • start_time: Start time in second.
  • end_time: End time in second.
  • speaker: Speaker tagging in string "speaker<N>"

Source Datasets

err_source_text: This is the erroneous ASR-Diarization results to be fixed. Has dev, eval folders ref_annotated_text: This is the human annotated ground-truth for evaluation. Only dev split is included.

  • Training Sources:

    • dev: 222 sessions
  • Development Sources:

    • dev: 13 sessions
  • Evaluation Sources:

    • eval: 11 Sessions

Access

The dataset can be accessed and downloaded through the HuggingFace Datasets library (i.e., This Repository).

Evaluation

This dataset can be evaluated by MeetEval Software

From PyPI

pip install meeteval

From source

git clone https://github.com/fgnt/meeteval
pip install -e ./meeteval

Evaluate the corrected segLST files:

python -m meeteval.wer cpwer -h err_source_text/dev/session_ge1nse2c.json -r ref_annotate_text/dev/session_ge1nse2c.json

Or after installation, you can use the following command alternatively.

meeteval-wer cpwer -h err_source_text/dev/session_ge1nse2c.json -r ref_annotate_text/dev/session_ge1nse2c.json

References

@inproceedings{park2024enhancing,
  title={Enhancing speaker diarization with large language models: A contextual beam search approach},
  author={Park, Tae Jin and Dhawan, Kunal and Koluguri, Nithin and Balam, Jagadeesh},
  booktitle={ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages={10861--10865},
  year={2024},
  organization={IEEE}
}
@InProceedings{MeetEval23,
  title={MeetEval: A Toolkit for Computation of Word Error Rates for Meeting Transcription Systems},
  author={von Neumann, Thilo and Boeddeker, Christoph and Delcroix, Marc and Haeb-Umbach, Reinhold},
  booktitle={CHiME-2023 Workshop, Dublin, England},
  year={2023}
}
Downloads last month
43
Edit dataset card