IPC / README.md
wnkh's picture
Update README.md
0443b10 verified
metadata
dataset_info:
  features:
    - name: image
      dtype: image
    - name: src_text
      dtype: string
    - name: translated_text
      dtype: string
    - name: shelfmark
      dtype: string
  splits:
    - name: train
      num_bytes: 151532201
      num_examples: 1383
  download_size: 172197795
  dataset_size: 151532201
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Interpres Parallel Corpus (IPC) is the first-of-it-kinds dataset that contains Medieval Latin image line, transcribed text, and expert translation. Image lines and corresponding texts are collected from the HTRomance project, expert translations are collected from Perseus Digital Library and Project Gutenberg and matched with lines by the chapter/book number and with the help of Gemini 3 Pro.