README.md

metadata

dataset_info:
  features:
    - name: section
      dtype: string
    - name: filename
      dtype: string
    - name: text
      dtype: string
    - name: detected_lang
      dtype: string
  splits:
    - name: train
      num_bytes: 7615427354
      num_examples: 14615
  download_size: 3638779542
  dataset_size: 7615427354
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc-by-4.0
task_categories:
  - text-generation
  - fill-mask
language:
  - en

An initial version of the Survivor Library OCR'd via rolmOCR.

• OCR of the books was completed by page, then joined together. You can split the text col on \f to get the pages for any given book
• OCR of ancient books is not perfect - some pages have issues. We plan to attempt to detect these and fix/update over time
• More details to come

Stats

Token Count

GPT-4 tiktoken token count:

        token_count
count  1.461500e+04
mean   1.228607e+05
std    2.448000e+05
min    7.000000e+00
25%    2.464200e+04
50%    4.321000e+04
75%    1.246360e+05
max    3.727261e+06

• Total count: 1795.61 M tokens