Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,197 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- af
|
5 |
+
- am
|
6 |
+
- ar
|
7 |
+
- as
|
8 |
+
- az
|
9 |
+
- be
|
10 |
+
- bg
|
11 |
+
- bn
|
12 |
+
- br
|
13 |
+
- bs
|
14 |
+
- ca
|
15 |
+
- cs
|
16 |
+
- cy
|
17 |
+
- da
|
18 |
+
- de
|
19 |
+
- en
|
20 |
+
- el
|
21 |
+
- eo
|
22 |
+
- es
|
23 |
+
- et
|
24 |
+
- eu
|
25 |
+
- fa
|
26 |
+
- fi
|
27 |
+
- fr
|
28 |
+
- fy
|
29 |
+
- ga
|
30 |
+
- gd
|
31 |
+
- gl
|
32 |
+
- gu
|
33 |
+
- ha
|
34 |
+
- he
|
35 |
+
- hi
|
36 |
+
- hr
|
37 |
+
- hu
|
38 |
+
- hy
|
39 |
+
- id
|
40 |
+
- is
|
41 |
+
- it
|
42 |
+
- ja
|
43 |
+
- jv
|
44 |
+
- ka
|
45 |
+
- kk
|
46 |
+
- km
|
47 |
+
- kn
|
48 |
+
- ko
|
49 |
+
- ku
|
50 |
+
- ky
|
51 |
+
- la
|
52 |
+
- lo
|
53 |
+
- lt
|
54 |
+
- lv
|
55 |
+
- mg
|
56 |
+
- mk
|
57 |
+
- ml
|
58 |
+
- mn
|
59 |
+
- mr
|
60 |
+
- ms
|
61 |
+
- my
|
62 |
+
- ne
|
63 |
+
- nl
|
64 |
+
- nb
|
65 |
+
- om
|
66 |
+
- or
|
67 |
+
- pa
|
68 |
+
- pl
|
69 |
+
- ps
|
70 |
+
- pt
|
71 |
+
- ro
|
72 |
+
- ru
|
73 |
+
- sa
|
74 |
+
- sd
|
75 |
+
- si
|
76 |
+
- sk
|
77 |
+
- sl
|
78 |
+
- so
|
79 |
+
- sq
|
80 |
+
- sr
|
81 |
+
- su
|
82 |
+
- sv
|
83 |
+
- sw
|
84 |
+
- ta
|
85 |
+
- te
|
86 |
+
- th
|
87 |
+
- tl
|
88 |
+
- tr
|
89 |
+
- ug
|
90 |
+
- uk
|
91 |
+
- ur
|
92 |
+
- uz
|
93 |
+
- vi
|
94 |
+
- xh
|
95 |
+
- yi
|
96 |
+
- zh
|
97 |
+
size_categories:
|
98 |
+
- 100M<n<1B
|
99 |
---
|
100 |
+
|
101 |
+
# Dataset Card for EntityCS
|
102 |
+
|
103 |
+
## Dataset Description
|
104 |
+
|
105 |
+
- Repository: https://github.com/huawei-noah/noah-research/tree/master/NLP/EntityCS
|
106 |
+
- Paper: https://aclanthology.org/2022.findings-emnlp.499.pdf
|
107 |
+
- Point of Contact: efstathia.christopoulou@huawei.com
|
108 |
+
|
109 |
+
### Dataset Summary
|
110 |
+
|
111 |
+
We use the English Wikipedia and leverage entity information from Wikidata to construct an entity-based Code Switching corpus.
|
112 |
+
To achieve this, we make use of wikilinks in Wikipedia, i.e. links from one page to another.
|
113 |
+
We use the English [Wikipedia dump](https://dumps.wikimedia.org/enwiki/latest/) (November 2021) and extract raw text with [WikiExtractor](https://github.com/attardi/wikiextractor) while keeping track of wikilinks.
|
114 |
+
Since we are interested in creating entity-level CS instances, we only keep sentences containing at least one wikilink.
|
115 |
+
Given an English sentence with wikilinks, we first map the entity in each wikilink to its corresponding Wikidata ID and
|
116 |
+
retrieve its available translations from Wikidata.
|
117 |
+
For each sentence, we check which languages have translations for all entities in that sentence, and consider those as candidates for code-switching.
|
118 |
+
We ensure all entities are code-switched to the same target language in a single sentence, avoiding noise from including too many languages.
|
119 |
+
To control the size of the corpus, we generate up to five code-switched sentences for each English sentence.
|
120 |
+
In particular, if fewer than five languages have translations available for all the entities in a sentence, we create code-switched instances with all of them.
|
121 |
+
Otherwise, we randomly select five target languages from the candidates.
|
122 |
+
If no candidate languages can be found, we do not code-switch the sentence, instead, we keep it as part of the English corpus.
|
123 |
+
Finally, we surround each entity with entity indicators (`<e>`, `</e>`).
|
124 |
+
|
125 |
+
### Supported Tasks and Leaderboards
|
126 |
+
|
127 |
+
The dataset was developped for intermediate pre-training of language models and can be used on any downstream task.
|
128 |
+
In the paper it's effectiveness is proven on entity-centric tasks, such as NER.
|
129 |
+
|
130 |
+
### Languages
|
131 |
+
|
132 |
+
The dataset covers 93 languages in total, including English.
|
133 |
+
|
134 |
+
## Dataset Structure
|
135 |
+
|
136 |
+
### Data Statistics
|
137 |
+
|
138 |
+
| Statistic | Count |
|
139 |
+
|:------------------------------|------------:|
|
140 |
+
| Languages | 93 |
|
141 |
+
| English Sentences | 54,469,214 |
|
142 |
+
| English Entities | 104,593,076 |
|
143 |
+
| Average Sentence Length | 23.37 |
|
144 |
+
| Average Entities per Sentence | 2 |
|
145 |
+
| CS Sentences per EN Sentence | ≤ 5 |
|
146 |
+
| CS Sentences | 231,124,422 |
|
147 |
+
| CS Entities | 420,907,878 |
|
148 |
+
|
149 |
+
### Data Fields
|
150 |
+
|
151 |
+
Each instance contains 3 fields:
|
152 |
+
- id: Unique ID of each sentence
|
153 |
+
- language: The language of choice for entity code-switching of the given sentence
|
154 |
+
- en_sentence: The original English sentence
|
155 |
+
- cs_sentence: The code-switched sentence
|
156 |
+
|
157 |
+
An example of what a data instance looks like:
|
158 |
+
```
|
159 |
+
{
|
160 |
+
'id': 19,
|
161 |
+
'en_sentence': 'The subs then enter a <en>coral reef</en> with many bright reflective colors.',
|
162 |
+
'cs_sentence': 'The subs then enter a <de>Korallenriff</de> with many bright reflective colors.',
|
163 |
+
'language': 'de'
|
164 |
+
}
|
165 |
+
```
|
166 |
+
|
167 |
+
### Data Splits
|
168 |
+
|
169 |
+
There is a single data split for each language. You can randomly select a few examples to serve as validation set.
|
170 |
+
|
171 |
+
|
172 |
+
### Limitations
|
173 |
+
|
174 |
+
An important limitation of the work is that before code-switching an entity, its morphological inflection is not checked.
|
175 |
+
This can lead to potential errors as the form of the CS entity might not agree with the surrounding context (e.g. plural).
|
176 |
+
There should be few cases as such, as we are only switching entities. However, this should be improved in a later version of the corpus.
|
177 |
+
Secondly, the diversity of languages used to construct the EntityCS corpus is restricted to the overlap between the available languages in WikiData and XLM-R pre-training.
|
178 |
+
This choice was for a better comparison between models, however it is possible to extend the corpus with more languages that XLM-R does not cover, following
|
179 |
+
the procedure presented in the paper.
|
180 |
+
|
181 |
+
### Citation
|
182 |
+
|
183 |
+
```html
|
184 |
+
@inproceedings{whitehouse-etal-2022-entitycs,
|
185 |
+
title = "{E}ntity{CS}: Improving Zero-Shot Cross-lingual Transfer with Entity-Centric Code Switching",
|
186 |
+
author = "Whitehouse, Chenxi and
|
187 |
+
Christopoulou, Fenia and
|
188 |
+
Iacobacci, Ignacio",
|
189 |
+
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2022",
|
190 |
+
month = dec,
|
191 |
+
year = "2022",
|
192 |
+
address = "Abu Dhabi, United Arab Emirates",
|
193 |
+
publisher = "Association for Computational Linguistics",
|
194 |
+
url = "https://aclanthology.org/2022.findings-emnlp.499",
|
195 |
+
pages = "6698--6714"
|
196 |
+
}
|
197 |
+
```
|