Datasets:

Modalities:
Audio
Text
Formats:
parquet
Languages:
English
ArXiv:
Libraries:
Datasets
Dask
License:
ylacombe HF staff commited on
Commit
9fb79a1
1 Parent(s): 964bfb1

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +61 -0
README.md CHANGED
@@ -23,4 +23,65 @@ configs:
23
  data_files:
24
  - split: train
25
  path: read/train-*
 
 
 
 
26
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
23
  data_files:
24
  - split: train
25
  path: read/train-*
26
+ license: cc-by-nc-4.0
27
+ language:
28
+ - en
29
+ pretty_name: The Expresso Dataset
30
  ---
31
+
32
+ # The Expresso Dataset
33
+ [[paper]](https://arxiv.org/abs/2308.05725) [[demo samples]](https://speechbot.github.io/expresso/) [[Original repository]](https://github.com/facebookresearch/textlesslib/tree/main/examples/expresso/dataset)
34
+
35
+ ## Introduction
36
+ The Expresso dataset is a high-quality (48kHz) expressive speech dataset that includes both expressively rendered read speech (8 styles, in mono wav format) and improvised dialogues (26 styles, in stereo wav format). The dataset includes 4 speakers (2 males, 2 females), and totals 40 hours (11h read, 30h improvised). The transcriptions of the read speech are also provided.
37
+
38
+ You can listen to samples from the Expresso Dataset at [this website](https://speechbot.github.io/expresso/).
39
+
40
+ ## Data Statistics
41
+ Here are the statistics of Expresso’s expressive styles:
42
+
43
+ ----------------------------------------------------------------
44
+ Style | Read (min) | Improvised (min) | total (hrs)
45
+ ------------------|------------|------------------|-------------
46
+ angry | - | 82 | 1.4
47
+ animal | - | 27 | 0.4
48
+ animal_directed | - | 32 | 0.5
49
+ awe | - | 92 | 1.5
50
+ bored | - | 92 | 1.5
51
+ calm | - | 93 | 1.6
52
+ child | - | 28 | 0.4
53
+ child_directed | - | 38 | 0.6
54
+ confused | 94 | 66 | 2.7
55
+ default | 133 | 158 | 4.9
56
+ desire | - | 92 | 1.5
57
+ disgusted | - | 118 | 2.0
58
+ enunciated | 116 | 62 | 3.0
59
+ fast | - | 98 | 1.6
60
+ fearful | - | 98 | 1.6
61
+ happy | 74 | 92 | 2.8
62
+ laughing | 94 | 103 | 3.3
63
+ narration | 21 | 76 | 1.6
64
+ non_verbal | - | 32 | 0.5
65
+ projected | - | 94 | 1.6
66
+ sad | 81 | 101 | 3.0
67
+ sarcastic | - | 106 | 1.8
68
+ singing* | - | 4 | .07
69
+ sleepy | - | 93 | 1.5
70
+ sympathetic | - | 100 | 1.7
71
+ whisper | 79 | 86 | 2.8
72
+ **Total** | **11.5h** | **34.4h** | **45.9h**
73
+ ----------------------------------------------------------------
74
+ *singing is the only improvised style that is not in dialogue format.
75
+
76
+ ## Audio Quality
77
+ The audio was recorded in a professional recording studio with minimal background noise at 48kHz/24bit. The files for read speech and singing are in a mono wav format; and for the dialog section in stereo (one channel per actor), where the original flow of turn-taking is preserved.
78
+
79
+
80
+ ### Read Speech
81
+ The `read` config contains all the read speech and singing style.
82
+
83
+ ## License
84
+ The Expresso dataset is distributed under the [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) license.
85
+
86
+ ## Reference
87
+ For more information, see the paper: [EXPRESSO: A Benchmark and Analysis of Discrete Expressive Speech Resynthesis](https://arxiv.org/abs/2308.05725), Tu Anh Nguyen*, Wei-Ning Hsu*, Antony D'Avirro*, Bowen Shi*, Itai Gat, Maryam Fazel-Zarani, Tal Remez, Jade Copet, Gabriel Synnaeve, Michael Hassid, Felix Kreuk, Yossi Adi⁺, Emmanuel Dupoux⁺, INTERSPEECH 2023.