File size: 844 Bytes
6769413
 
b009740
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6769413
b92f0d8
 
 
32e55ac
b92f0d8
 
 
 
 
 
 
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
---
license: odc-by
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: text
    dtype: string
  - name: timestamp
    dtype: int64
  - name: url
    dtype: string
  - name: score
    dtype: int64
  splits:
  - name: train
    num_bytes: 5152049
    num_examples: 501
  download_size: 2669350
  dataset_size: 5152049
---


このデータセットはmC4の日本語データに対し、文の割合に応じて1~5段階で人手で評価したものです。  
文の割合はデータセットの```score```に格納しています。
1. 文の割合が20%以下
2. 文の割合が20~40%
3. 文の割合が40~60%
4. 文の割合が60~80%
5. 文の割合が80~100%

データ量は500件と少ないですが、mC4のゴミデータ削除に役立てればと思います。