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

Composed of documents from:

- biglam/hmd_newspapers
- biglam/lampeter_corpus
- TheBritishLibrary/blbooks
- dell-research-harvard/AmericanStories
- dell-research-harvard/NewsWire
- storytracer/LoC-PD-Books
- some books from https://ota.ox.ac.uk/

The perfect score is larger than 0 (zero). The score is composed of how far the metrics are from 100,
either smaller of larger than 100, except for quality which can be larger than 0.

# Short documents

Filtered documents smaller than 32k characters, English, pre-1900.
```
|
| Scanned rows: 13,078,094
|
| Sources (8):
| - American: 8,620,056
| - HMD: 2,231,598
| - British: 1,952,414
| - NewsWire: 149,029
| - Brix: 118,927
| - LOC-PD: 6,029
| - Lampeter: 36
| - Brit: 5
|
| Metric       |    Mean |     Min |     P5 |     P95 |      Max |
|--------------|---------|---------|--------|---------|----------|
| Char length  | 2,134.4 |    64.0 |  176.0 | 7,668.0 | 31,245.0 |
| Unique chars |   47.70 |    8.00 |  34.00 |   61.00 |    64.00 |
| Token count  |   383.7 |     7.0 |   32.0 | 1,364.0 |  6,463.0 |
| Sentences    |   20.21 |    2.00 |   2.00 |   74.00 | 2,742.00 |
| Quality      |   98.21 |   25.00 |  86.90 |  104.24 |   113.78 |
| Compression  |  110.94 |   56.56 |  97.07 |  130.14 |   152.73 |
| Entropy      |  106.16 |   56.74 |  98.49 |  114.61 |   126.83 |
| Dict hit     |   88.75 |   25.00 |  73.00 |   97.45 |   100.00 |
| Alpha        |   94.78 |   25.00 |  83.78 |   99.82 |   100.00 |
| Vowel        |   96.12 |   25.42 |  87.50 |  100.00 |   100.00 |
| ASCII        |   98.66 |   25.04 |  94.80 |  100.00 |   100.00 |
| Score        |  -38.11 | -100.00 | -77.42 |  -11.59 |    17.12 |
```

# Long documents

Filtered large documents, smaller than 10mil characters, English, pre-1900.
```
|
| Rows: 198,833
|
| Sources (5):
| - Brix: 79,302
| - LOC: 57,924
| - HMD: 51,050
| - OTA: 10,473
| - Lampeter: 84
|
| Metric       |      Mean |      Min |       P5 |       P95 |         Max |
|--------------|-----------|----------|----------|-----------|-------------|
| Char length  | 254,605.2 | 19,101.0 | 32,783.6 | 769,263.4 | 8,465,466.0 |
| Unique chars |     63.54 |    44.00 |    62.00 |     64.00 |       64.00 |
| Token count  |  46,145.9 |  3,369.0 |  5,870.0 | 140,146.8 | 1,834,218.0 |
| Sentences    |  2,732.34 |    11.00 |   220.00 |  7,748.00 |  253,533.00 |
| Quality      |     96.65 |    25.15 |    88.27 |    101.60 |      104.81 |
| Compression  |     93.62 |    60.18 |    85.62 |    102.58 |      114.30 |
| Entropy      |    107.11 |    73.54 |   101.03 |    113.90 |      122.83 |
| Dict hit     |     91.74 |    27.65 |    81.71 |     97.49 |       99.66 |
| Alpha        |     95.29 |    25.20 |    88.93 |     99.27 |      100.00 |
| Vowel        |     97.61 |    43.64 |    94.25 |     99.75 |      100.00 |
| ASCII        |     97.95 |    15.65 |    93.00 |     99.84 |      100.00 |
| Score        |    -28.23 |   -99.86 |   -45.55 |    -15.99 |       -6.64 |
```

# Gutenberg

Filtered Gutenberg books, English, pre-1900, split into chunks (approx chapters).
```
|
| Rows: 240,752
|
| Metric       |     Mean |    Min |      P5 |      P95 |         Max |
|--------------|----------|--------|---------|----------|-------------|
| Char length  | 25,174.6 |  318.0 | 1,052.0 | 78,397.9 | 4,580,626.0 |
| Unique chars |    53.08 |  11.00 |   42.00 |    63.00 |       64.00 |
| Token count  |  4,580.1 |   46.0 |   173.0 | 14,079.9 |   901,007.0 |
| Sentences    |   195.37 |   2.00 |    9.00 |   589.00 |   87,528.00 |
| Quality      |    97.93 |  50.02 |   84.84 |   103.40 |      110.35 |
| Compression  |    93.98 |  59.30 |   86.61 |   103.04 |      121.26 |
| Entropy      |   100.75 |  52.34 |   96.96 |   107.03 |      127.63 |
| Dict hit     |    95.87 |  26.33 |   89.14 |    99.20 |      100.00 |
| Alpha        |    98.06 |  27.64 |   91.44 |   100.00 |      100.00 |
| Vowel        |    99.42 |  43.59 |   97.81 |   100.00 |      100.00 |
| ASCII        |    98.42 |  26.84 |   92.91 |   100.00 |      100.00 |
| Score        |   -15.37 | -99.85 |  -32.81 |    -5.42 |       15.96 |
```

## Citation

If you find this dataset valuable, please consider citing:

```bibtex
@misc{vintage-v1,
  title  = {Vintage v1},
  author = {Cristi Constantin},
  month  = {June},
  year   = {2026},
  url = {https://huggingface.co/datasets/croqaz/vintage-v1}
}
```