File size: 2,255 Bytes
4051e24
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f1601a3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
dataset_info:
  features:
  - name: region
    dtype: string
  - name: province_code
    dtype: int64
  - name: province_name
    dtype: string
  - name: filename
    dtype: string
  - name: text
    dtype: string
  - name: speakerID
    dtype: string
  - name: gender
    dtype: int64
  - name: audio
    dtype: audio
  splits:
  - name: train
    num_bytes: 51014515813.0
    num_examples: 15023
  - name: test
    num_bytes: 6669046477.0
    num_examples: 2026
  - name: valid
    num_bytes: 6334942186.0
    num_examples: 1900
  download_size: 59828302626
  dataset_size: 64018504476.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
  - split: valid
    path: data/valid-*
---

Multi-Dialect Vietnamese: Task, Dataset, Baseline Models and Challenges

## Dataset Overview 

Overall, our dataset offers a comprehensive representation of Vietnamese dialects, including 63 provincial dialects. It consists of 102.56 hours of audio recordings, with nearly 19,000 records ob-
tained from 12,955 speakers. The accompanying transcripts consist of over 1.2 million words, with a distinct vocabulary of 5,155 unique words. The train, validation, and test sets were split in an 8:1:1 ratio based on duration, resulting in 81.43 hours,
10.26 hours, and 10.87 hours, respectively. This ratio extended to the number of records, speakers, and words as well. Notwithstanding such a ratio, the unique word counts in the validation set
(2,660 unique words) and test set (2,723 unique words) does not differ significantly from from the training set (4,813 unique words), thus preserving the vocabulary diversity. Across the 63 provincial
dialects, with the exception of the number of speakers exhibiting imbalance, the remaining attributes duration, number of records, words, and unique words – are well-balanced

## Citation Information


```
@article{Multi-Dialect Vietnamese: Task, Dataset, Baseline Models and Challenges,
  title={Multi-Dialect Vietnamese: Task, Dataset, Baseline Models and Challenges},
  author={Nguyen Van Dinh, Thanh Chi Dang, Luan Thanh Nguyen, Kiet Van Nguyen},
  year=2024,
  month=October},
  url={https://huggingface.co/datasets/nguyendv02/ViMD_Dataset}
```