Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Overview
|
| 2 |
+
This dataset is a subset of the huggingface wikipedia dataset with ~70'000 rows, each about a person on wikipedia.
|
| 3 |
+
Each row contains the original wikipedia texts as sentences,
|
| 4 |
+
as well as a paraphrased version of each sentence. For both versions full texts with the entity the wikipedia page is about being masked.
|
| 5 |
+
|
| 6 |
+
# features
|
| 7 |
+
- id: the id in the original dataset
|
| 8 |
+
- url: the link to the wikipedia page
|
| 9 |
+
- title: the title of the wikipedia page
|
| 10 |
+
- text: the original wikipedia text
|
| 11 |
+
- sentences: text split to sentences
|
| 12 |
+
- paraphrased_sentences: text split to sentences, with each sentence paraphrased (e.g. mutated a bit)
|
| 13 |
+
- masked_text_original: original text with entity masked in every occurence (<mask> as token)
|
| 14 |
+
- masked_entities_original: array of entities masked in masked_text_original
|
| 15 |
+
- masked_text_paraphrased: paraphrased text with entity masked in every occurence
|
| 16 |
+
- masked_entities_paraphrased: array of entities msked in masked_text_paraphrased
|
| 17 |
+
|
| 18 |
---
|
| 19 |
+
annotations_creators:
|
| 20 |
+
- no-annotation
|
| 21 |
+
- machine-generated
|
| 22 |
+
language:
|
| 23 |
+
- en
|
| 24 |
+
language_creators:
|
| 25 |
+
- found
|
| 26 |
+
license:
|
| 27 |
+
- afl-3.0
|
| 28 |
+
multilinguality:
|
| 29 |
+
- monolingual
|
| 30 |
+
pretty_name: wikipedia persons paraphrased and masked
|
| 31 |
+
size_categories:
|
| 32 |
+
- 10K<n<100K
|
| 33 |
+
source_datasets:
|
| 34 |
+
- extended|wikipedia
|
| 35 |
+
tags: []
|
| 36 |
+
task_categories:
|
| 37 |
+
- fill-mask
|
| 38 |
+
task_ids:
|
| 39 |
+
- slot-filling
|