README.md

metadata

license: mit
task_categories:
  - feature-extraction
language:
  - ru
  - en
dataset_info:
  features:
    - name: task
      list:
        - name: content
          dtype: string
        - name: role
          dtype: string
  splits:
    - name: train
      num_bytes: 129602.9798657718
      num_examples: 119
    - name: test
      num_bytes: 32673.020134228187
      num_examples: 30
  download_size: 41756
  dataset_size: 162276
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*

The Central Bank of Russia (CBR) posts decisions on the registration of securities on its website. We are interested in bonds. The dataset consists of: instructions for the model (in English) plus the text of the decision (in Russian) from the CBR and a dictionary with keys and values from the text of the decisions

UPD, Dec 1, 2024: Structure of the dataset has been changed:

• 'train' and 'test' splits (80%/20%);
• the task is to extract only securities numbers from the provided text;
• output -> string (not a dictionary);
• provided text is shorter
• duplicates deleted