Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,326 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
|
3 |
+
# Dataset Card for Text2Tech Curated Documents
|
4 |
+
|
5 |
+
## Dataset Summary
|
6 |
+
|
7 |
+
This dataset is the result of converting a UIMA CAS 0.4 JSON export from the Inception annotation tool into a simplified format suitable for Natural Language Processing tasks. Specifically, it provides configurations for Named Entity Recognition (NER), Entity Linking (EL), and Relation Extraction (RE).
|
8 |
+
|
9 |
+
The conversion process utilized the `dkpro-cassis` library to load the original annotations and `spaCy` for tokenization. The final dataset is structured similarly to the DFKI-SLT/mobie dataset to ensure compatibility and ease of use with the Hugging Face ecosystem.
|
10 |
+
|
11 |
+
This version of the dataset loader provides configurations for:
|
12 |
+
|
13 |
+
* **Named Entity Recognition (ner)**: NER tags use spaCy's BILUO tagging scheme.
|
14 |
+
* **Entity Linking (el)**: Entity mentions are linked to external knowledge bases.
|
15 |
+
* **Relation Extraction (re)**: Relations between entities are annotated.
|
16 |
+
|
17 |
+
## Supported Tasks and Leaderboards
|
18 |
+
|
19 |
+
* **Tasks**: Named Entity Recognition, Entity Linking, Relation Extraction
|
20 |
+
* **Leaderboards**: More Information Needed
|
21 |
+
|
22 |
+
## Languages
|
23 |
+
|
24 |
+
The text in the dataset is in English.
|
25 |
+
|
26 |
+
## Dataset Structure
|
27 |
+
|
28 |
+
### Data Instances
|
29 |
+
|
30 |
+
#### ner
|
31 |
+
|
32 |
+
An example of 'train' looks as follows.
|
33 |
+
|
34 |
+
```json
|
35 |
+
{
|
36 |
+
"docid": "138",
|
37 |
+
"tokens": [
|
38 |
+
"\"",
|
39 |
+
"Samsung",
|
40 |
+
"takes",
|
41 |
+
"aim",
|
42 |
+
"at",
|
43 |
+
"blood",
|
44 |
+
"pressure",
|
45 |
+
"monitoring",
|
46 |
+
"with",
|
47 |
+
"the",
|
48 |
+
"Galaxy",
|
49 |
+
"Watch",
|
50 |
+
"Active",
|
51 |
+
"..."
|
52 |
+
],
|
53 |
+
"ner_tags": [
|
54 |
+
0,
|
55 |
+
1,
|
56 |
+
0,
|
57 |
+
0,
|
58 |
+
0,
|
59 |
+
2,
|
60 |
+
3,
|
61 |
+
4,
|
62 |
+
0,
|
63 |
+
0,
|
64 |
+
5,
|
65 |
+
6,
|
66 |
+
7,
|
67 |
+
"..."
|
68 |
+
]
|
69 |
+
}
|
70 |
+
```
|
71 |
+
|
72 |
+
#### el
|
73 |
+
|
74 |
+
An example of 'train' looks as follows.
|
75 |
+
|
76 |
+
```json
|
77 |
+
{
|
78 |
+
"docid": "138",
|
79 |
+
"tokens": [
|
80 |
+
"\"",
|
81 |
+
"Samsung",
|
82 |
+
"takes",
|
83 |
+
"aim",
|
84 |
+
"at",
|
85 |
+
"blood",
|
86 |
+
"pressure",
|
87 |
+
"monitoring",
|
88 |
+
"with",
|
89 |
+
"the",
|
90 |
+
"Galaxy",
|
91 |
+
"Watch",
|
92 |
+
"Active",
|
93 |
+
"..."
|
94 |
+
],
|
95 |
+
"ner_tags": [
|
96 |
+
0,
|
97 |
+
1,
|
98 |
+
0,
|
99 |
+
0,
|
100 |
+
0,
|
101 |
+
2,
|
102 |
+
3,
|
103 |
+
4,
|
104 |
+
0,
|
105 |
+
0,
|
106 |
+
5,
|
107 |
+
6,
|
108 |
+
7,
|
109 |
+
"..."
|
110 |
+
],
|
111 |
+
|
112 |
+
"entity_mentions": [
|
113 |
+
{
|
114 |
+
"text": "Samsung",
|
115 |
+
"start": 1,
|
116 |
+
"end": 2,
|
117 |
+
"char_start": 1,
|
118 |
+
"char_end": 8,
|
119 |
+
"type": 0,
|
120 |
+
"entity_id": "http://www.wikidata.org/entity/Q124989916"
|
121 |
+
},
|
122 |
+
"..."
|
123 |
+
]
|
124 |
+
}
|
125 |
+
```
|
126 |
+
|
127 |
+
#### re
|
128 |
+
|
129 |
+
An example of 'train' looks as follows.
|
130 |
+
|
131 |
+
```json
|
132 |
+
{
|
133 |
+
"docid": "138",
|
134 |
+
"tokens": [
|
135 |
+
"\"",
|
136 |
+
"Samsung",
|
137 |
+
"takes",
|
138 |
+
"aim",
|
139 |
+
"at",
|
140 |
+
"blood",
|
141 |
+
"pressure",
|
142 |
+
"monitoring",
|
143 |
+
"with",
|
144 |
+
"the",
|
145 |
+
"Galaxy",
|
146 |
+
"Watch",
|
147 |
+
"Active",
|
148 |
+
"..."
|
149 |
+
],
|
150 |
+
"ner_tags": [
|
151 |
+
0,
|
152 |
+
1,
|
153 |
+
0,
|
154 |
+
0,
|
155 |
+
0,
|
156 |
+
2,
|
157 |
+
3,
|
158 |
+
4,
|
159 |
+
0,
|
160 |
+
0,
|
161 |
+
5,
|
162 |
+
6,
|
163 |
+
7,
|
164 |
+
"..."
|
165 |
+
],
|
166 |
+
"relations": [
|
167 |
+
{
|
168 |
+
"id": "138-0",
|
169 |
+
"head_start": 706,
|
170 |
+
"head_end": 708,
|
171 |
+
"head_type": 2,
|
172 |
+
"tail_start": 706,
|
173 |
+
"tail_end": 708,
|
174 |
+
"tail_type": 2,
|
175 |
+
"type": 0
|
176 |
+
},
|
177 |
+
"..."
|
178 |
+
]
|
179 |
+
}
|
180 |
+
```
|
181 |
+
|
182 |
+
### Data Fields
|
183 |
+
|
184 |
+
#### ner
|
185 |
+
|
186 |
+
* `docid`: A `string` feature representing the document identifier.
|
187 |
+
* `tokens`: A `list` of `string` features representing the tokens in the document.
|
188 |
+
* `ner_tags`: A `list` of classification labels using spaCy's BILUO tagging scheme. The mapping from ID to tag is as follows:
|
189 |
+
|
190 |
+
**BILUO Tagging Scheme:**
|
191 |
+
- **B-** (Begin): First token of a multi-token entity
|
192 |
+
- **I-** (Inside): Inner tokens of a multi-token entity
|
193 |
+
- **L-** (Last): Final token of a multi-token entity
|
194 |
+
- **U-** (Unit): Single token entity
|
195 |
+
- **O** (Outside): Non-entity token
|
196 |
+
|
197 |
+
```json
|
198 |
+
{
|
199 |
+
"O": 0,
|
200 |
+
"U-Organization": 1,
|
201 |
+
"B-Method": 2,
|
202 |
+
"I-Method": 3,
|
203 |
+
"L-Method": 4,
|
204 |
+
"B-Technological System": 5,
|
205 |
+
"I-Technological System": 6,
|
206 |
+
"L-Technological System": 7,
|
207 |
+
"U-Technological System": 8,
|
208 |
+
"U-Method": 9,
|
209 |
+
"B-Material": 10,
|
210 |
+
"L-Material": 11,
|
211 |
+
"I-Material": 12,
|
212 |
+
"B-Organization": 13,
|
213 |
+
"L-Organization": 14,
|
214 |
+
"I-Organization": 15,
|
215 |
+
"U-Material": 16,
|
216 |
+
"B-Technical Field": 17,
|
217 |
+
"L-Technical Field": 18,
|
218 |
+
"I-Technical Field": 19,
|
219 |
+
"U-Technical Field": 20
|
220 |
+
}
|
221 |
+
```
|
222 |
+
|
223 |
+
#### el
|
224 |
+
|
225 |
+
* `docid`: A `string` feature representing the document identifier.
|
226 |
+
* `tokens`: A `list` of `string` features representing the tokens in the document.
|
227 |
+
* `entity_mentions`: A `list` of `struct` features containing:
|
228 |
+
* `text`: a `string` feature.
|
229 |
+
* `start`: token offset start, a `int32` feature.
|
230 |
+
* `end`: token offset end, a `int32` feature.
|
231 |
+
* `char_start`: character offset start, a `int32` feature.
|
232 |
+
* `char_end`: character offset end, a `int32` feature.
|
233 |
+
* `type`: a classification label. The mapping from ID to entity type is as follows:
|
234 |
+
|
235 |
+
```json
|
236 |
+
{
|
237 |
+
"Organization": 0,
|
238 |
+
"Method": 1,
|
239 |
+
"Technological System": 2,
|
240 |
+
"Material": 3,
|
241 |
+
"Technical Field": 4
|
242 |
+
}
|
243 |
+
```
|
244 |
+
|
245 |
+
* `entity_id`: a `string` feature representing the entity identifier from a knowledge base.
|
246 |
+
|
247 |
+
#### re
|
248 |
+
|
249 |
+
* `docid`: A `string` feature representing the document identifier.
|
250 |
+
* `tokens`: A `list` of `string` features representing the tokens in the document.
|
251 |
+
* `ner_tags`: A `list` of classification labels, corresponding to the NER task.
|
252 |
+
* `relations`: A `list` of `struct` features containing:
|
253 |
+
* `id`: a `string` feature representing the relation identifier.
|
254 |
+
* `head_start`: token offset start of the head entity, an `int32` feature.
|
255 |
+
* `head_end`: token offset end of the head entity, an `int32` feature.
|
256 |
+
* `head_type`: a classification label for the head entity type.
|
257 |
+
* `tail_start`: token offset start of the tail entity, an `int32` feature.
|
258 |
+
* `tail_end`: token offset end of the tail entity, an `int32` feature.
|
259 |
+
* `tail_type`: a classification label for the tail entity type.
|
260 |
+
* `type`: a classification label for the relation type. The mapping from ID to relation type is as follows:
|
261 |
+
|
262 |
+
```json
|
263 |
+
{
|
264 |
+
"ts:executes": 0,
|
265 |
+
"org:develops_or_provides": 1,
|
266 |
+
"ts:contains": 2,
|
267 |
+
"ts:made_of": 3,
|
268 |
+
"ts:uses": 4,
|
269 |
+
"ts:supports": 5,
|
270 |
+
"met:employs": 6,
|
271 |
+
"met:processes": 7,
|
272 |
+
"mat:transformed_to": 8,
|
273 |
+
"org:collaborates": 9,
|
274 |
+
"met:creates": 10,
|
275 |
+
"met:applied_to": 11,
|
276 |
+
"ts:processes": 12
|
277 |
+
}
|
278 |
+
```
|
279 |
+
|
280 |
+
### Data Splits
|
281 |
+
|
282 |
+
Please add information about your data splits here. For example:
|
283 |
+
|
284 |
+
* **train**: X samples
|
285 |
+
* **validation**: Y samples
|
286 |
+
* **test**: Z samples
|
287 |
+
|
288 |
+
## Dataset Creation
|
289 |
+
|
290 |
+
The dataset was created by converting JSON files exported from the Inception annotation tool. The `inception_converter.py` script was used to process these files. This script uses the `dkpro-cassis` library to load the UIMA CAS JSON data and `spaCy` for tokenization and creating BIO tags for the NER task. The data was then split into three separate files for NER, EL, and RE tasks.
|
291 |
+
|
292 |
+
## Considerations for Using the Data
|
293 |
+
|
294 |
+
### Social Impact of Dataset
|
295 |
+
|
296 |
+
More Information Needed
|
297 |
+
|
298 |
+
### Discussion of Biases
|
299 |
+
|
300 |
+
More Information Needed
|
301 |
+
|
302 |
+
### Other Known Limitations
|
303 |
+
|
304 |
+
More Information Needed
|
305 |
+
|
306 |
+
## Additional Information
|
307 |
+
|
308 |
+
### Dataset Curators
|
309 |
+
|
310 |
+
Amir Safari
|
311 |
+
|
312 |
+
### Licensing Information
|
313 |
+
|
314 |
+
Please specify the license for this dataset.
|
315 |
+
|
316 |
+
### Citation Information
|
317 |
+
|
318 |
+
Please provide a BibTeX citation for your dataset.
|
319 |
+
|
320 |
+
```bibtex
|
321 |
+
author = {Amir Safari},
|
322 |
+
title = {Text2Tech Curated Documents},
|
323 |
+
year = {2025},
|
324 |
+
publisher = {Hugging Face}
|
325 |
+
}
|
326 |
+
```
|