---
dataset_info:
  features:
  - name: sentences
    list:
    - name: id
      dtype: int64
    - name: speaker
      dtype: 'null'
    - name: text
      dtype: string
    - name: tokens
      list:
      - name: id
        dtype: int64
      - name: lemma
        dtype: string
      - name: text
        dtype: string
      - name: xpos
        dtype: string
  - name: coref_chains
    sequence:
      sequence:
        sequence: int64
  - name: id
    dtype: string
  - name: text
    dtype: string
  - name: genre
    dtype: string
  - name: meta_data
    struct:
    - name: comment
      dtype: string
  splits:
  - name: train
    num_bytes: 36047013
    num_examples: 1345
  - name: validation
    num_bytes: 3639179
    num_examples: 135
  - name: test
    num_bytes: 3703845
    num_examples: 207
  download_size: 11763612
  dataset_size: 43390037
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
---

This dataset was generated by reformatting [`coref-data/korean_ecmt_raw`](https://huggingface.co/datasets/coref-data/korean_ecmt_raw) into the indiscrim coreference format. See that repo for dataset details.

See [ianporada/coref-data](https://github.com/ianporada/coref-data) for additional conversion details and the conversion script.

Please create an issue in the repo above or in this dataset repo for any questions.