r52 / README.md
joao-luz's picture
Update README.md
1106a31 verified
metadata
dataset_info:
  features:
    - name: label
      dtype:
        class_label:
          names:
            '0': acq
            '1': alum
            '2': bop
            '3': carcass
            '4': cocoa
            '5': coffee
            '6': copper
            '7': cotton
            '8': cpi
            '9': cpu
            '10': crude
            '11': dlr
            '12': earn
            '13': fuel
            '14': gas
            '15': gnp
            '16': gold
            '17': grain
            '18': heat
            '19': housing
            '20': income
            '21': instal-debt
            '22': interest
            '23': ipi
            '24': iron-steel
            '25': jet
            '26': jobs
            '27': lead
            '28': lei
            '29': livestock
            '30': lumber
            '31': meal-feed
            '32': money-fx
            '33': money-supply
            '34': nat-gas
            '35': nickel
            '36': orange
            '37': pet-chem
            '38': platinum
            '39': potato
            '40': reserves
            '41': retail
            '42': rubber
            '43': ship
            '44': strategic-metal
            '45': sugar
            '46': tea
            '47': tin
            '48': trade
            '49': veg-oil
            '50': wpi
            '51': zinc
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 4317997
      num_examples: 6532
    - name: test
      num_bytes: 1537041
      num_examples: 2568
  download_size: 3055131
  dataset_size: 5855038
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*

This is the R52 dataset, a subset of the Reuters 21587 dataset categorized in 52 distinct topics.

The data was obtained from https://github.com/yao8839836/text_gcn, keeping this repo's train and test splits. The structure was adapted into the Dataset format using the following ClassLabel mapping:

Label Name
0 acq
1 alum
2 bop
3 carcass
4 cocoa
5 coffee
6 copper
7 cotton
8 cpi
9 cpu
10 crude
11 dlr
12 earn
13 fuel
14 gas
15 gnp
16 gold
17 grain
18 heat
19 housing
20 income
21 instal-debt
22 interest
23 ipi
24 iron-steel
25 jet
26 jobs
27 lead
28 lei
29 livestock
30 lumber
31 meal-feed
32 money-fx
33 money-supply
34 nat-gas
35 nickel
36 orange
37 pet-chem
38 platinum
39 potato
40 reserves
41 retail
42 rubber
43 ship
44 strategic-metal
45 sugar
46 tea
47 tin
48 trade
49 veg-oil
50 wpi
51 zinc