Hugging Face's logo

Datasets:
tartuNLP
/
EstNER

Tasks:
Token Classification
Modalities:
Text
Formats:
parquet
Languages:
Estonian
Size:
10K - 100K
Libraries:
Datasets
Dask
License:
Dataset card Viewer Files Community
1
adorkin commited on
Commit
e61389f
·
verified ·
1 Parent(s): a841612

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +213 -1
README.md CHANGED
@@ -30,4 +30,216 @@ language:
30
  pretty_name: EstNER
31
  size_categories:
32
  - 10K<n<100K
33
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
30
  pretty_name: EstNER
31
  size_categories:
32
  - 10K<n<100K
33
+ ---
34
+
35
+ # Dataset Card for EstNER
36
+
37
+ ## Dataset Description
38
+
39
+ EstNER dataset for named entity recogintion in Estonian language comprised of two parts: _New EstNER_ and _Reannotated EstNER_ (refer to the corresponding sections of this readme for additional details).
40
+
41
+ By default the joint version of the dataset is loaded.
42
+
43
+ ```python
44
+ from datasets import load_dataset
45
+
46
+ ds = load_dataset("tartuNLP/EstNER")
47
+ ```
48
+
49
+ Each part can be loaded individually, as well.
50
+
51
+ ```python
52
+ from datasets import load_dataset
53
+
54
+ new_ds = load_dataset("tartuNLP/EstNER", "estner-new")
55
+ reannotated_ds = load_dataset("tartuNLP/EstNER", "estner-reannotated")
56
+ ```
57
+
58
+
59
+ ### New Estonian NER dataset
60
+
61
+ The dataset is a collection of Estonian news and social media texts annotated with named entities.
62
+
63
+ #### Dataset statistics
64
+
65
+ The dataset is divided into training, development and test sets. The annotations can be hierarchical, meaning that there can be one named entity inside another. The maximum number of levels in the hierarchical annotations is three.
66
+
67
+ | | Train | Dev | Test | Total |
68
+ |-----------------|--------|-------|-------|--------|
69
+ | Documents | 78 | 16 | 15 | 109 |
70
+ | Sentences | 7001 | 882 | 890 | 8773 |
71
+ | Tokens | 111858 | 13130 | 14686 | 139674 |
72
+ |1-level entities | 7480 | 497 | 938 | 8915 |
73
+ |2-level entities | 571 | 44 | 59 | 674 |
74
+ |3-level entities | 27 | 0 | 1 | 28 |
75
+
76
+ #### Annotated entities
77
+
78
+ The dataset is annotated with the following entities:
79
+ * PER - person names
80
+ * GPE - geopolitical entities
81
+ * LOC - geographical locations
82
+ * ORG - organizations
83
+ * PROD - products, things, works of art
84
+ * EVENT - events
85
+ * DATE - dates
86
+ * TIME - times
87
+ * TITLE - titles and professions
88
+ * MONEY - monetary expressions
89
+ * PERCENT - percentages
90
+
91
+ ##### Level 1 entities
92
+
93
+ | | Train | Dev | Test | Total |
94
+ |---------|-------|-----|-------|-------|
95
+ | PER | 2601 | 109 | 299 | 3009 |
96
+ | ORG | 1177 | 85 | 150 | 1412 |
97
+ | LOC | 449 | 31 | 35 | 515 |
98
+ | GPE | 1253 | 129 | 231 | 1613 |
99
+ | TITLE | 702 | 19 | 59 | 772 |
100
+ | PROD | 624 | 60 | 117 | 801 |
101
+ | EVENT | 230 | 15 | 26 | 271 |
102
+ | DATE | 746 | 64 | 77 | 887 |
103
+ | TIME | 103 | 6 | 6 | 115 |
104
+ | PERCENT | 75 | 11 | 1 | 87 |
105
+ | MONEY | 118 | 12 | 1 | 131 |
106
+ | Total | 8078 | 541 | 994 | 9613 |
107
+
108
+ ##### Level 2 entities
109
+
110
+ | | Train | Dev | Test | Total |
111
+ |---------|-------|-----|-------|-------|
112
+ | PER | 108 | 1 | 14 | 123 |
113
+ | ORG | 92 | 8 | 6 | 106 |
114
+ | LOC | 25 | 1 | 0 | 26 |
115
+ | GPE | 379 | 35 | 42 | 456 |
116
+ | TITLE | 3 | 0 | 0 | 3 |
117
+ | PROD | 4 | 0 | 0 | 4 |
118
+ | EVENT | 1 | 0 | 0 | 1 |
119
+ | DATE | 10 | 0 | 0 | 10 |
120
+ | TIME | 0 | 0 | 0 | 0 |
121
+ | PERCENT | 0 | 0 | 0 | 0 |
122
+ | MONEY | 0 | 0 | 0 | 0 |
123
+ | Total | 622 | 45 | 62 | 729 |
124
+
125
+
126
+ ##### Level 3 entities
127
+
128
+ | | Train | Dev | Test | Total |
129
+ |---------|-------|-----|-------|-------|
130
+ | PER | 1 | 0 | 0 | 1 |
131
+ | ORG | 0 | 0 | 0 | 0 |
132
+ | LOC | 1 | 0 | 0 | 1 |
133
+ | GPE | 25 | 0 | 1 | 26 |
134
+ | TITLE | 0 | 0 | 0 | 0 |
135
+ | PROD | 0 | 0 | 0 | 0 |
136
+ | EVENT | 0 | 0 | 0 | 0 |
137
+ | DATE | 0 | 0 | 0 | 0 |
138
+ | TIME | 0 | 0 | 0 | 0 |
139
+ | PERCENT | 0 | 0 | 0 | 0 |
140
+ | MONEY | 0 | 0 | 0 | 0 |
141
+ | Total | 27 | 0 | 1 | 28 |
142
+
143
+
144
+ ### Reannotated Estonian NER dataset
145
+
146
+ This is the Estonian NER dataset ([Tkachenko, 2010](https://core.ac.uk/download/pdf/16270382.pdf); [Tkachenko et al., 2013](https://aclanthology.org/W13-2412.pdf)) reannotated with a richer set of entities.
147
+
148
+ #### Dataset statistics
149
+
150
+ The dataset is divided into training, development and test sets. The annotations can be hierarchical, meaning that there can be one named entity inside another. The maximum number of levels in the hierarchical annotations is three.
151
+
152
+ | | Train | Dev | Test | Total |
153
+ |-----------------|--------|-------|-------|--------|
154
+ | Documents | 525 | 18 | 39 | 582 |
155
+ | Sentences | 9965 | 2415 | 1907 | 14287 |
156
+ | Tokens | 155983 | 32890 | 28370 | 217243 |
157
+ |1-level entities | 13918 | 2571 | 2396 | 18885 |
158
+ |2-level entities | 987 | 223 | 122 | 1332 |
159
+ |3-level entities | 40 | 14 | 4 | 58 |
160
+
161
+ #### Annotated entities
162
+
163
+ Originally, the Estonian NER dataset was annotated with PER, ORG and LOC entities only. The reannotated version is annotated with the following entities:
164
+ * PER - person names
165
+ * GPE - geopolitical entities
166
+ * LOC - geographical locations
167
+ * ORG - organizations
168
+ * PROD - products, things, works of art
169
+ * EVENT - events
170
+ * DATE - dates
171
+ * TIME - times
172
+ * TITLE - titles and professions
173
+ * MONEY - monetary expressions
174
+ * PERCENT - percentages
175
+
176
+ ##### Level 1 entities
177
+
178
+ | | Train | Dev | Test | Total |
179
+ |---------|-------|-----|-------|-------|
180
+ | PER | 3563 | 642 | 722 | 4927 |
181
+ | ORG | 3215 | 504 | 541 | 4260 |
182
+ | LOC | 328 | 118 | 61 | 507 |
183
+ | GPE | 3377 | 714 | 479 | 4570 |
184
+ | TITLE | 1302 | 171 | 209 | 1682 |
185
+ | PROD | 874 | 161 | 66 | 1101 |
186
+ | EVENT | 56 | 13 | 17 | 86 |
187
+ | DATE | 1346 | 308 | 186 | 1840 |
188
+ | TIME | 456 | 39 | 30 | 525 |
189
+ | PERCENT | 137 | 62 | 58 | 257 |
190
+ | MONEY | 291 | 76 | 153 | 520 |
191
+ | Total | 14945 | 2808| 2522 | 20275 |
192
+
193
+ ##### Level 2 entities
194
+
195
+ | | Train | Dev | Test | Total |
196
+ |---------|-------|-----|-------|-------|
197
+ | PER | 46 | 7 | 4 | 57 |
198
+ | ORG | 180 | 31 | 12 | 223 |
199
+ | LOC | 58 | 12 | 8 | 78 |
200
+ | GPE | 745 | 160 | 101 | 1006 |
201
+ | TITLE | 6 | 0 | 0 | 6 |
202
+ | PROD | 3 | 0 | 0 | 3 |
203
+ | EVENT | 5 | 0 | 0 | 5 |
204
+ | DATE | 7 | 34 | 1 | 42 |
205
+ | TIME | 0 | 0 | 0 | 0 |
206
+ | PERCENT | 1 | 0 | 0 | 1 |
207
+ | MONEY | 0 | 0 | 0 | 0 |
208
+ | Total | 1051 | 126 | 244 | 1421 |
209
+
210
+
211
+ ##### Level 3 entities
212
+
213
+ | | Train | Dev | Test | Total |
214
+ |---------|-------|-----|-------|-------|
215
+ | PER | 1 | 0 | 0 | 1 |
216
+ | ORG | 1 | 0 | 0 | 0 |
217
+ | LOC | 0 | 1 | 0 | 1 |
218
+ | GPE | 38 | 13 | 4 | 26 |
219
+ | TITLE | 0 | 0 | 0 | 0 |
220
+ | PROD | 0 | 0 | 0 | 0 |
221
+ | EVENT | 0 | 0 | 0 | 0 |
222
+ | DATE | 0 | 0 | 0 | 0 |
223
+ | TIME | 0 | 0 | 0 | 0 |
224
+ | PERCENT | 0 | 0 | 0 | 0 |
225
+ | MONEY | 0 | 0 | 0 | 0 |
226
+ | Total | 40 | 14 | 4 | 58 |
227
+
228
+ ## BibTeX entry and citation info
229
+
230
+ ```
231
+ @inproceedings{sirts-2023-estonian,
232
+ title = "{E}stonian Named Entity Recognition: New Datasets and Models",
233
+ author = "Sirts, Kairit",
234
+ editor = {Alum{\"a}e, Tanel and
235
+ Fishel, Mark},
236
+ booktitle = "Proceedings of the 24th Nordic Conference on Computational Linguistics (NoDaLiDa)",
237
+ month = may,
238
+ year = "2023",
239
+ address = "T{\'o}rshavn, Faroe Islands",
240
+ publisher = "University of Tartu Library",
241
+ url = "https://aclanthology.org/2023.nodalida-1.76",
242
+ pages = "752--761",
243
+ abstract = "This paper presents the annotation process of two Estonian named entity recognition (NER) datasets, involving the creation of annotation guidelines for labeling eleven different types of entities. In addition to the commonly annotated entities such as person names, organization names, and locations, the annotation scheme encompasses geopolitical entities, product names, titles/roles, events, dates, times, monetary values, and percents. The annotation was performed on two datasets, one involving reannotating an existing NER dataset primarily composed of news texts and the other incorporating new texts from news and social media domains. Transformer-based models were trained on these annotated datasets to establish baseline predictive performance. Our findings indicate that the best results were achieved by training a single model on the combined dataset, suggesting that the domain differences between the datasets are relatively small.",
244
+ }
245
+ ```