File size: 3,469 Bytes
392f3f3
a4e3e9f
 
 
 
 
f048662
 
a747124
a4e3e9f
a747124
 
da97246
c703bc9
 
 
97e5238
a747124
97e5238
 
 
 
 
 
 
 
 
a747124
97e5238
 
c32ee38
97e5238
 
ca6e169
 
 
 
 
e9090dc
ca6e169
 
c3278a9
 
 
 
 
 
 
 
 
 
 
 
 
 
a6162dd
 
 
 
 
 
 
 
 
c3278a9
 
5eb009a
7faa766
 
038dac6
7faa766
038dac6
7faa766
 
 
 
405f496
 
106faea
 
279b006
106faea
405f496
 
7faa766
 
ca6e169
97e5238
 
c31bca3
 
97e5238
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105

---
license: creativeml-openrail-m
language:
- en
- ku
size_categories:
- 10K<n<100K
pretty_name: "Kurdish TED (KUTED)"
task_categories:
- audio-to-audio
- automatic-speech-recognition
tags:
  - TED
  - S2TT
  - KURDISH
  - SPEECH_Translation
---


#  Table of Contents

1. [Dataset Description](#data_description)
2. [Uses](#uses)
5. [Evaluation](#evaluation)
7. [Citation](#citation)

## Kurdish TED (KUTED)
Kurdish TED (KUTED) is the first Speech-to-Text-Translation (S2TT) dataset for Central Kurdish language derived from TED Talks and TEDx. 
The corpus consists of 91,000 pairs, encompassing 170 hours of English audio, 1.65 million English tokens, and 1.40 million Central Kurdish tokens.
This dataset is evaluated on speech E2E S2TT, Cascaded S2TT and T2TT tasks. 


## Uses
Kurdish TED can be used for the followin tasks:
- Speech-to-Text-Translation (En->CKB)
- Speech-to-Speech-Translation (En->CKB)
- Text-to-Text-Translation (En->CKB and CKB->EN)
- Auotomatic Speech Recognition (En)


## How to Get Started with KUTED

## 1) Download
Using this code you can download all parts of KUTED dataset:

```python
from datasets import load_dataset

dataset = load_dataset('aranemini/kurdishted')

train_dataset = load_dataset('aranemini/kurdishted', split='train')
test_dataset = load_dataset('aranemini/kurdishted', split='test')
test_dataset = load_dataset('aranemini/kurdishted', split='validation')

```
OR


You can clone this dataset repository directly using `git`:

```bash
git lfs install
git clone https://huggingface.co/datasets/aranemini/kurdishted
```

Then, Go to the download directory and extract tar files.

## 3) Data format
For each utterance, there is a folder. The name of the folder corresponds to the ID of each record. Inside each folder, you will find the id.wav file and a JSON file in the following format.

In each JSON file, there are two versions of the Kurdish transcription. The kurdish_raw field refers to the transcription provided by TED Translators, while the kurdish_standardized field is the normalized and standardized version of the Kurdish transcription. For more details on the applied normalization and standardization process, please refer to the reference paper.

Here is a sample of the dataset in JSON format:

```json
{
 "english": "In the 19th and 20th centuries, people worshipped the nation, the Aryan race, the communist state.",
"id": "clean_how_we_can_face_the_future_without_fear_together_captions_segment_12",
 "kurdish_raw": "‫له‌ سه‌ده‌ی 19 و 20،‬ ‫،خه‌ڵك نه‌ته‌وه‌یان ده‌په‌رست‬ ‫.ڕه‌گه‌زی ئاری، ده‌وڵه‌تی كۆمۆنستی‬",
 "kurdish_standardized": ". لە سەدەی ١٩ و ٢٠ ، خەڵک نەتەوەیان دەپەرست ، ڕەگەزی ئاری ، دەوڵەتی کۆمۆنیستی",
 "part": "test",
 "path": "test/clean_how_we_can_face_the_future_without_fear_together_captions_segment_12/clean_how_we_can_face_the_future_without_fear_together_captions_segment_12.wav"
 }
```


# Citation

If you are using this dataset in your reserach please cite the following reference:

**BibTeX:**

```bibtex
@article{kuted2024,
  author  = {Aran Emini, Josep Crego, Daban Q. Jaff, Antoine Laurent},
  title   = {Central Kurdish Speech Translation: Corpus Creation, Evaluation, and Orthographic Standardization},
  journal = {},
  year    = {2024},
  volume  = {},
  number  = {},
  pages   = {},
  url     = {}
}
```