README.md

metadata

dataset_info:
  features:
    - name: docid
      dtype: string
    - name: previous_text
      dtype: string
    - name: gold_text
      dtype: string
    - name: citations
      sequence:
        sequence: string
    - name: oracle_documents_passages
      sequence:
        sequence: string
    - name: top_10_passages
      sequence: string
  splits:
    - name: train
      num_bytes: 1367048300
      num_examples: 5000
    - name: test
      num_bytes: 277656017
      num_examples: 1000
  download_size: 635009461
  dataset_size: 1644704317
configs:
  - config_name: default
    data_files:
      - split: train
        path: bm25_relevant_passages_oracle_documents/train-*
      - split: test
        path: bm25_relevant_passages_oracle_documents/test-*
  - config_name: data
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
  - config_name: BRPOD
    data_files:
      - split: train
        path: bm25_relevant_passages_oracle_documents/train-*
      - split: test
        path: bm25_relevant_passages_oracle_documents/test-*
  - config_name: DRPOD
    data_files:
      - split: train
        path: >-
          dense_relevant_passages_oracle_documents/jhu-clsp_BERT-DPR-CLERC-ft/train-*
      - split: test
        path: >-
          dense_relevant_passages_oracle_documents/jhu-clsp_BERT-DPR-CLERC-ft/test-*
task_categories:
  - text-generation
language:
  - en
pretty_name: CLERC Generation Workshop