lmoncla commited on
Commit
25988b2
·
verified ·
1 Parent(s): ba7f3b2

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +28 -17
README.md CHANGED
@@ -8,11 +8,21 @@ widget:
8
  * ALBI, (Géog.) ville de France, capitale de l'Albigeois, dans le haut
9
  Languedoc : elle est sur le Tarn. Long. 19. 49. lat. 43. 55. 44.
10
  - text: >-
11
- HILPERHAUSEN, (Géog.) ville d'Allemagne en Franconie, sur la Werra, au comté de Henneberg, entre Cobourg & Smalcalde ; elle appartient à une branche de la maison de Saxe-Gotha. Long. 28. 15. lat. 50. 35. (D. J.)
 
 
 
 
 
 
 
 
 
 
12
  ---
13
 
14
 
15
- # bert-base-french-cased-edda-ner
16
 
17
 
18
  <!-- Provide a quick summary of what the model is/does. -->
@@ -72,7 +82,7 @@ import torch
72
  from datasets import load_dataset
73
 
74
 
75
- pipe = pipeline("token-classification", model="GEODE/bert-base-french-cased-edda-ner", aggregation_strategy="simple", device=device)
76
 
77
  content = "* ALBI, (Géog.) ville de France, capitale de l'Albigeois, dans le haut Languedoc : elle est sur le Tarn. Long. 19. 49. lat. 43. 55. 44."
78
 
@@ -197,28 +207,29 @@ For full training details and results please see the GitHub repository: [https:/
197
  ### Evaluation
198
 
199
 
200
- * Overall model performances (Test set)
201
 
202
 
203
  | | Precision | Recall | F-score |
204
  |---|:---:|:---:|:---:|
205
- | | 90.1 | 93.7 | 91.9 |
206
 
207
 
208
 
209
- * Model performances by entity (Test set)
210
 
211
- | | Precision | Recall | F-score |
212
  |---|:---:|:---:|:---:|
213
- | NC-Spatial | 91.6 | 95.3 | 93.4 |
214
- | NP-Spatial | 95.9 | 95.5 | 95.7 |
215
- | Relation | 89.4 | 94.7 | 91.9 |
216
- | Latlong | 98.1 | 96.8 | 97.4 |
217
- | NC-Person | 67.5 | 84.0 | 74.9 |
218
- | NP-Person | 87.4 | 89.2 | 88.3 |
219
- | NP-Misc | 72.4 | 76.6 | 74.4 |
220
- | Head | 97.6 | 97.2 | 97.4 |
221
- | Domain-mark | 99.2 | 100.0 | 99.6 |
 
222
 
223
 
224
 
@@ -228,4 +239,4 @@ For full training details and results please see the GitHub repository: [https:/
228
 
229
 
230
  The authors are grateful to the [ASLAN project](https://aslan.universite-lyon.fr) (ANR-10-LABX-0081) of the Université de Lyon, for its financial support within the French program "Investments for the Future" operated by the National Research Agency (ANR).
231
- Data courtesy the [ARTFL Encyclopédie Project](https://artfl-project.uchicago.edu), University of Chicago.
 
8
  * ALBI, (Géog.) ville de France, capitale de l'Albigeois, dans le haut
9
  Languedoc : elle est sur le Tarn. Long. 19. 49. lat. 43. 55. 44.
10
  - text: >-
11
+ HILPERHAUSEN, (Géog.) ville d'Allemagne en Franconie, sur la Werra, au comté
12
+ de Henneberg, entre Cobourg & Smalcalde ; elle appartient à une branche de
13
+ la maison de Saxe-Gotha. Long. 28. 15. lat. 50. 35. (D. J.)
14
+ datasets:
15
+ - GEODE/GeoEDdA
16
+ metrics:
17
+ - f1
18
+ - precision
19
+ - recall
20
+ base_model:
21
+ - almanach/camembert-base
22
  ---
23
 
24
 
25
+ # camembert-base-edda-span-classification
26
 
27
 
28
  <!-- Provide a quick summary of what the model is/does. -->
 
82
  from datasets import load_dataset
83
 
84
 
85
+ pipe = pipeline("token-classification", model="GEODE/camembert-base-edda-span-classification", aggregation_strategy="simple", device=device)
86
 
87
  content = "* ALBI, (Géog.) ville de France, capitale de l'Albigeois, dans le haut Languedoc : elle est sur le Tarn. Long. 19. 49. lat. 43. 55. 44."
88
 
 
207
  ### Evaluation
208
 
209
 
210
+ * Overall micro-average model performances (token-based)
211
 
212
 
213
  | | Precision | Recall | F-score |
214
  |---|:---:|:---:|:---:|
215
+ | | 91.5 | 94.8 | 93.1 |
216
 
217
 
218
 
219
+ * Token-based model performances (Test set)
220
 
221
+ | | Precision | Recall | F-score | Support
222
  |---|:---:|:---:|:---:|
223
+ | NC-Spatial | 96.7 | 95.1 | 95.9 | 592
224
+ | NP-Spatial | 95.9 | 95.5 | 95.7 | 717
225
+ | Relation | 89.8 | 95.6 | 92.6 | 452
226
+ | Latlong | 97.0 | 98.5 | 97.7 | 789
227
+ | NC-Person | 70.4 | 78.4 | 74.2 | 222
228
+ | NP-Person | 88.6 | 90.4 | 89.5 | 198
229
+ | NP-Misc | 69.0 | 82.9 | 75.3 | 175
230
+ | Head | 97.3 | 98.0 | 97.6 | 254
231
+ | Domain-mark | 99.0 | 100.0 | 99.5 | 392
232
+
233
 
234
 
235
 
 
239
 
240
 
241
  The authors are grateful to the [ASLAN project](https://aslan.universite-lyon.fr) (ANR-10-LABX-0081) of the Université de Lyon, for its financial support within the French program "Investments for the Future" operated by the National Research Agency (ANR).
242
+ Data courtesy the [ARTFL Encyclopédie Project](https://artfl-project.uchicago.edu), University of Chicago.