|
--- |
|
language: |
|
- nl |
|
dataset_info: |
|
features: |
|
- name: tokens |
|
sequence: string |
|
- name: pos |
|
sequence: string |
|
- name: lemma |
|
sequence: string |
|
- name: mw_id |
|
sequence: string |
|
- name: corpus |
|
dtype: string |
|
splits: |
|
- name: train |
|
num_bytes: 14145953 |
|
num_examples: 10812 |
|
- name: validation |
|
num_bytes: 2231572 |
|
num_examples: 1686 |
|
- name: test |
|
num_bytes: 1931275 |
|
num_examples: 1639 |
|
download_size: 3290324 |
|
dataset_size: 18308800 |
|
configs: |
|
- config_name: default |
|
data_files: |
|
- split: train |
|
path: data/train-* |
|
- split: validation |
|
path: data/validation-* |
|
- split: test |
|
path: data/test-* |
|
--- |
|
|
|
# Galahad training data |
|
|
|
As taken from [Github](https://github.com/INL/galahad-corpus-data/tree/1.0.1/training-data). |
|
|
|
## Statistics |
|
|
|
The directory [statistics/](https://huggingface.co/datasets/ivdnt/galahad-corpus-data/tree/main/statistics) contains some frequency calculations such as a frequency list of all lemmata and part-of-speechs. Note that multi-words did not get any special treatment in the data structure, so both for lemma and PoS you may see concatenated labels such as `lemma1+lemma2` as a single label for a given token. |
|
|
|
### Text and token counts |
|
|
|
Total: 14,137 texts, 390,534 tokens |
|
|
|
<details> |
|
<summary>Texts and tokens per corpus</summary> |
|
|
|
- clvn: 857 texts, 27,654 tokens |
|
- couranten: 800 texts, 29,577 tokens |
|
- dbnl-excerpts-15: 138 texts, 9,611 tokens |
|
- dbnl-excerpts-16: 797 texts, 10,002 tokens |
|
- dbnl-excerpts-17: 256 texts, 11,626 tokens |
|
- dbnl-excerpts-18: 212 texts, 9,986 tokens |
|
- dbnl-excerpts-19: 503 texts, 15,301 tokens |
|
- dictionary-quotations-15: 2,231 texts, 41,012 tokens |
|
- dictionary-quotations-16: 1,826 texts, 45,851 tokens |
|
- dictionary-quotations-17: 1,901 texts, 45,836 tokens |
|
- dictionary-quotations-18: 1,756 texts, 46,182 tokens |
|
- dictionary-quotations-19: 1,540 texts, 34,740 tokens |
|
- letters-as-loot: 1,320 texts, 63,156 tokens |
|
|
|
</details> |
|
|
