README.md

metadata

dataset_info:
  - config_name: chunked-4096
    features:
      - name: id
        dtype: string
      - name: text
        dtype: string
      - name: source
        dtype: string
      - name: added
        dtype: timestamp[us]
      - name: created
        dtype: timestamp[us]
      - name: metadata
        struct:
          - name: Source-File
            dtype: string
          - name: olmocr-version
            dtype: string
          - name: pdf-total-pages
            dtype: int64
          - name: total-fallback-pages
            dtype: int64
          - name: total-input-tokens
            dtype: int64
          - name: total-output-tokens
            dtype: int64
      - name: attributes
        struct:
          - name: pdf_page_numbers
            sequence:
              sequence: int64
      - name: chunk_id
        dtype: int64
      - name: original_id
        dtype: string
    splits:
      - name: train
        num_bytes: 240028507
        num_examples: 14754
    download_size: 88286371
    dataset_size: 240028507
  - config_name: default
    features:
      - name: id
        dtype: string
      - name: text
        dtype: string
      - name: source
        dtype: string
      - name: added
        dtype: timestamp[s]
      - name: created
        dtype: timestamp[s]
      - name: metadata
        struct:
          - name: Source-File
            dtype: string
          - name: olmocr-version
            dtype: string
          - name: pdf-total-pages
            dtype: int64
          - name: total-input-tokens
            dtype: int64
          - name: total-output-tokens
            dtype: int64
          - name: total-fallback-pages
            dtype: int64
      - name: attributes
        struct:
          - name: pdf_page_numbers
            sequence:
              sequence: int64
    splits:
      - name: train
        num_bytes: 178667251.98773006
        num_examples: 4228
    download_size: 79820485
    dataset_size: 178667251.98773006
configs:
  - config_name: chunked-4096
    data_files:
      - split: train
        path: chunked-4096/train-*
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
language:
  - en
task_categories:
  - text-generation
  - feature-extraction
tags:
  - ocr
  - document

NapierOne PDFs - converted with olmOCR

PDFs from NapierOne (see 'pdf-total' in the napierone aws bucket) converted to text via the olmOCR pipeline

• outputs with <= 100 chars and 'low quality pdfs' were filtered out
• the 4228 rows in default config represent approx 93,595 input PDF pages

Citation

@article{DAVIES2022301330,
  title = {NapierOne: A modern mixed file data set alternative to Govdocs1},
  journal = {Forensic Science International: Digital Investigation},
  volume = {40},
  pages = {301330},
  year = {2022},
  issn = {2666-2817},
  doi = {https://doi.org/10.1016/j.fsidi.2021.301330},
  url = {https://www.sciencedirect.com/science/article/pii/S2666281721002560},
  author = {Simon R. Davies and Richard Macfarlane and William J. Buchanan},
  keywords = {Corpus, Mixed file dataset, Govdocs1, Malware, Ransomware, Entropy, Forensics},
}