---
license: cc
language:
- en
pretty_name: Vintage-v1
tags:
- vintage
- historical
size_categories:
- 10K<n<100K
configs:
- config_name: default
  data_files:
  - split: train
    path: "data/train-*.parquet"
---

# Vintage dataset

Contains:

- CLMET v3.1 -- 335 books
- ECCO -- 3,102 books
- EEBO -- 34,937 books
- EVANS -- 5,013 books

All books are cleaned (best effort) from XML to Markdown.

Original sources:

- https://fedora.clarin-d.uni-saarland.de/clmet/clmet.html
- https://textpartnership.net/pages/faq.html
- https://dropbox.com/sh/inyy253jvytoxcu/AAAT9nMPo0sa5aXEgs5SC5aGa?dl=0 -- Evans TCP bulk files
- https://app.box.com/s/zj7pzfokxde4glrhebxsavbbxzyr3ogz -- Evans TCP bulk files
- https://dropbox.com/sh/inhwjphw682i2gf/AAC8NixNye8Gp0smYBTly2Y9a?dl=0 -- ECCO TCP bulk files
- https://app.box.com/s/6jbuf443i145f97c4t56z3garu3u2j09 -- ECCO TCP bulk files
- https://dropbox.com/sh/pfx619wnjdck2lj/AAAeQjd_dv29oPymNoKJWfEYa?dl=0 -- EEBO Phase 1 and 2 bulk files
- https://app.box.com/s/jjzmnrx98dkvanipopz3nxkvymnjccht -- EEBO Phase 1 and 2 bulk files

Where the dialogs were annotated, they are now extracted as JSON.

# Citation

If you find this dataset valuable, please consider citing:

```bibtex
@misc{vintage-v2,
  title  = {Vintage v2},
  author = {Cristi Constantin},
  month  = {July},
  year   = {2026},
  url = {https://huggingface.co/datasets/croqaz/vintage-v2}
}
```