|
--- |
|
license: cc0-1.0 |
|
size_categories: |
|
- 1M<n<10M |
|
source_datasets: |
|
- gbenson/webui-dom-snapshots |
|
task_categories: |
|
- text-classification |
|
pretty_name: WebUI tokens (unlabelled) |
|
dataset_info: |
|
features: |
|
- name: text |
|
dtype: string |
|
splits: |
|
- name: train |
|
num_bytes: 71634460 |
|
num_examples: 1616815 |
|
download_size: 69704391 |
|
dataset_size: 71634460 |
|
configs: |
|
- config_name: default |
|
data_files: |
|
- split: train |
|
path: data/train-* |
|
--- |
|
# Dataset Card for WebUI tokens (unlabelled) |
|
|
|
Every token over 5 characters long from [`gbenson/webui-dom-snapshots`](https://huggingface.co/datasets/gbenson/webui-dom-snapshots). |
|
|
|
- **Curated by:** [Gary Benson](https://gbenson.net/) |
|
- **License:** [CC0 1.0 Universal](https://creativecommons.org/publicdomain/zero/1.0/) |
|
|
|
## Uses |
|
|
|
I'm using it to develop a [DOM-aware tokenizer](https://github.com/gbenson/dom-tokenizers) for HTML. |
|
|
|
## Bias, Risks, and Limitations |
|
|
|
- 87% of the source dataset was English language websites, with no other language exceeding 2% of the total |
|
- Non-ASCII tokens have been coerced to ASCII using [Unidecode](https://pypi.org/project/Unidecode/) where the result appears visually similar |
