Dataset Viewer
Auto-converted to Parquet Duplicate
audio_rel
stringlengths
22
47
text
stringlengths
2
22
audio
listlengths
8k
32k
sampling_rate
int64
16k
16k
num_samples
int64
8k
32k
duration
float64
0.5
2
the_free_press/7264-92316-0025_000.wav
the_free_press
[ 0.00323486328125, 0.003997802734375, 0.003997802734375, 0.00390625, 0.00396728515625, 0.004241943359375, 0.004608154296875, 0.004638671875, 0.00457763671875, 0.0040283203125, 0.003753662109375, 0.0035400390625, 0.002960205078125, 0.002838134765625, 0.002532958984375, 0.002166748046875,...
16,000
10,880
0.68
the_free_press/7264-92316-0022_000.wav
the_free_press
[0.037261962890625,0.0382080078125,0.03955078125,0.0352783203125,0.028839111328125,0.02410888671875,(...TRUNCATED)
16,000
9,760
0.61
the_free_press/7264-92316-0011_001.wav
the_free_press
[-0.0048828125,-0.0010986328125,-0.001373291015625,-0.00140380859375,-0.000579833984375,0.0011291503(...TRUNCATED)
16,000
12,641
0.790063
the_free_press/7264-92316-0024_005.wav
the_free_press
[0.01031494140625,0.0086669921875,0.007843017578125,0.005584716796875,0.002716064453125,-0.000335693(...TRUNCATED)
16,000
12,640
0.79
the_free_press/7264-92316-0005_002.wav
the_free_press
[0.011627197265625,0.001861572265625,-0.002471923828125,-0.000091552734375,0.007781982421875,-0.0011(...TRUNCATED)
16,000
10,560
0.66
the_free_press/7264-92316-0005_007.wav
the_free_press
[-0.02972412109375,-0.03228759765625,-0.021392822265625,-0.012359619140625,-0.00799560546875,-0.0084(...TRUNCATED)
16,000
13,920
0.87
the_free_press/7264-92316-0028_003.wav
the_free_press
[0.000579833984375,0.00042724609375,0.000274658203125,-0.000091552734375,0.000091552734375,0.0002441(...TRUNCATED)
16,000
13,920
0.87
the_free_press/7264-92316-0038_000.wav
the_free_press
[0.060394287109375,0.057403564453125,0.05364990234375,0.052825927734375,0.05316162109375,0.051239013(...TRUNCATED)
16,000
10,080
0.63
the_free_press/7264-92316-0013_000.wav
the_free_press
[0.004425048828125,0.00921630859375,0.01458740234375,0.020111083984375,0.024688720703125,0.028045654(...TRUNCATED)
16,000
11,520
0.72
the_free_press/7264-92316-0021_001.wav
the_free_press
[-0.00146484375,-0.001373291015625,-0.001220703125,-0.001312255859375,-0.00146484375,-0.001617431640(...TRUNCATED)
16,000
12,160
0.76
End of preview. Expand in Data Studio

LibriPhrase-100

We construct LibriPhrase\footnote{\url{https://github.com/gusrud1103/LibriPhrase}} (LP) , retaining anchors with durations between 0.5–2 seconds. To study the effect of anchor-class size, the training set includes LP-100 ($\sim$12k classes) and LP-460 ($\sim$78k classes), with two sampled subsets from LP-460: LP-460-20k and LP-460-40k. GigaPhrase\footnote{\url{https://github.com/aizhiqi-work/GigaPhrase-1000}} is derived from the GigaSpeech corpus using the same MFA-based segmentation, providing a large collection of phrase-level anchors. Combining GP-1000 with LP-460 forms the LP–GP–1460 training set, containing approximately 155k phrase classes.

Downloads last month
8