|
|
--- |
|
|
dataset_info: |
|
|
features: |
|
|
- name: id |
|
|
dtype: string |
|
|
- name: title |
|
|
dtype: string |
|
|
- name: description |
|
|
dtype: string |
|
|
- name: languages_count |
|
|
dtype: int64 |
|
|
- name: wikidata |
|
|
dtype: string |
|
|
splits: |
|
|
- name: train |
|
|
num_bytes: 556636125 |
|
|
num_examples: 15783 |
|
|
download_size: 170502275 |
|
|
dataset_size: 556636125 |
|
|
configs: |
|
|
- config_name: default |
|
|
data_files: |
|
|
- split: train |
|
|
path: data/train-* |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
--- |
|
|
|
|
|
This dataset includes "things" from wikidata dump, where a thing is an item, concept, or an idea, which could be easily recognized and comprehended by an average English speaker. |
|
|
|
|
|
We made an effort to exclude anything related to politics, religion, as well as all geographical names to avoid potential bias. All potentially NSFW entries were also attempted to be excluded. |
|
|
|
|
|
We do not guarantee the accuracy of our filters and dataset safety, and admit that creation of this dataset is largely based on semi-arbitrary design decisions, which may be biased as to what counts as a "thing". |
