multi_news_RST / README.md
eeoonn's picture
Update README.md
994f5a9 verified
metadata
dataset_info:
  features:
    - name: document
      dtype: string
    - name: segmentation
      sequence: int64
    - name: parsing
      sequence: string
  splits:
    - name: train
      num_bytes: 1230225872
      num_examples: 44972
    - name: validation
      num_bytes: 150037757
      num_examples: 5622
    - name: test
      num_bytes: 159144540
      num_examples: 5622
  download_size: 619018558
  dataset_size: 1539408169
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
language:
  - en

Multi-News applied RST

RST: DMRST_parser

  1. "|||||" separates each document in the "document" field.
  2. "||" separates each EDU in the "document" field.
  3. The "parsing" field shows the relation of each EDU.

Example, [[edu.strip() for edu in docu.split("||") if edu.strip()] for docu in data["train"][0]["document"].split("|||||")]