Commit
·
522e605
1
Parent(s):
a12a6ae
Update README.md
Browse files
README.md
CHANGED
@@ -13,16 +13,21 @@ metrics:
|
|
13 |
- recall
|
14 |
- f1
|
15 |
widget:
|
16 |
-
- text:
|
|
|
17 |
נובל ל שלום, תמך בגלוי ב מועמדותו ל משרת ה מושל.
|
18 |
-
- text:
|
|
|
19 |
ל ידי רק ב ימים אלה.
|
20 |
-
- text:
|
21 |
-
ב
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
|
|
|
|
|
|
26 |
תעופה של היא.
|
27 |
pipeline_tag: token-classification
|
28 |
model-index:
|
@@ -45,6 +50,8 @@ model-index:
|
|
45 |
- type: recall
|
46 |
value: 0.7607296137339056
|
47 |
name: Recall
|
|
|
|
|
48 |
---
|
49 |
|
50 |
# SpanMarker
|
@@ -180,6 +187,51 @@ trainer.save_model("span_marker_model_id-finetuned")
|
|
180 |
| 1.3178 | 3000 | 0.0052 | 0.8322 | 0.7655 | 0.7975 | 0.9714 |
|
181 |
| 1.7571 | 4000 | 0.0053 | 0.8008 | 0.7735 | 0.7870 | 0.9712 |
|
182 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
183 |
### Framework Versions
|
184 |
- Python: 3.10.12
|
185 |
- SpanMarker: 1.5.0
|
@@ -190,13 +242,21 @@ trainer.save_model("span_marker_model_id-finetuned")
|
|
190 |
|
191 |
## Citation
|
192 |
|
193 |
-
|
194 |
```
|
195 |
-
@
|
196 |
-
author = {
|
197 |
-
|
198 |
-
|
199 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
200 |
}
|
201 |
```
|
202 |
|
|
|
13 |
- recall
|
14 |
- f1
|
15 |
widget:
|
16 |
+
- text: >-
|
17 |
+
אלי ויזל, פרופסור ב אוניברסיטת בוסטון, ש סילבר התאמץ הרבה למען זכייתו ב פרס
|
18 |
נובל ל שלום, תמך בגלוי ב מועמדותו ל משרת ה מושל.
|
19 |
+
- text: >-
|
20 |
+
מאמרו של תום שגב, " ה קרב על סן סימון היה או לא היה " (" ה ארץ " 105), הגיע
|
21 |
ל ידי רק ב ימים אלה.
|
22 |
+
- text: >-
|
23 |
+
רק ב דבריו של ה רב אברהם טולדאנו, משגיח ב ישיבת ה רעיון ה יהודי ו מספר 4 ב
|
24 |
+
רשימת כך ל ה כנסת, היו כבר הוראות מעשיות: " אלוקים ייקום דמו ו אנו ניקום את
|
25 |
+
הוא.
|
26 |
+
- text: >-
|
27 |
+
מרכז ה מידע ל זכויות ה אדם ב ה שטחים, " בצלם ", מפרסם מ פעם ל פעם דפי מידע ו
|
28 |
+
ב המ פרטים על ה נעשה ב ה שטחים ב תחומים שונים.
|
29 |
+
- text: >-
|
30 |
+
גרוסבורד נהג לבדו ב ה מכונית, ב דרכו מ ה עיר מיניאפוליס ב אינדיאנה ל נמל ה
|
31 |
תעופה של היא.
|
32 |
pipeline_tag: token-classification
|
33 |
model-index:
|
|
|
50 |
- type: recall
|
51 |
value: 0.7607296137339056
|
52 |
name: Recall
|
53 |
+
language:
|
54 |
+
- he
|
55 |
---
|
56 |
|
57 |
# SpanMarker
|
|
|
187 |
| 1.3178 | 3000 | 0.0052 | 0.8322 | 0.7655 | 0.7975 | 0.9714 |
|
188 |
| 1.7571 | 4000 | 0.0053 | 0.8008 | 0.7735 | 0.7870 | 0.9712 |
|
189 |
|
190 |
+
### Evaluation Results
|
191 |
+
|
192 |
+
| | precision | recall | f1 | number |
|
193 |
+
|:------------------------|------------:|------------:|------------:|-------------:|
|
194 |
+
| eval_loss | 0.00522302 | 0.00522302 | 0.00522302 | 0.00522302 |
|
195 |
+
| eval_ANG | 0 | 0 | 0 | 3 |
|
196 |
+
| eval_DUC | 0 | 0 | 0 | 2 |
|
197 |
+
| eval_EVE | 0 | 0 | 0 | 12 |
|
198 |
+
| eval_FAC | 0.333333 | 0.0833333 | 0.133333 | 12 |
|
199 |
+
| eval_GPE | 0.887931 | 0.85124 | 0.869198 | 121 |
|
200 |
+
| eval_LOC | 0.703704 | 0.678571 | 0.690909 | 28 |
|
201 |
+
| eval_ORG | 0.719298 | 0.689076 | 0.703863 | 119 |
|
202 |
+
| eval_PER | 0.889447 | 0.917098 | 0.903061 | 193 |
|
203 |
+
| eval_WOA | 0 | 0 | 0 | 9 |
|
204 |
+
| eval_overall_precision | 0.832244 | 0.832244 | 0.832244 | 0.832244 |
|
205 |
+
| eval_overall_recall | 0.765531 | 0.765531 | 0.765531 | 0.765531 |
|
206 |
+
| eval_overall_f1 | 0.797495 | 0.797495 | 0.797495 | 0.797495 |
|
207 |
+
| eval_overall_accuracy | 0.971418 | 0.971418 | 0.971418 | 0.971418 |
|
208 |
+
| eval_runtime | 34.3336 | 34.3336 | 34.3336 | 34.3336 |
|
209 |
+
| eval_samples_per_second | 23.505 | 23.505 | 23.505 | 23.505 |
|
210 |
+
| eval_steps_per_second | 11.767 | 11.767 | 11.767 | 11.767 |
|
211 |
+
| epoch | 2 | 2 | 2 | 2 |
|
212 |
+
|
213 |
+
### Tests Results
|
214 |
+
|
215 |
+
| | precision | recall | f1 | number |
|
216 |
+
|:------------------------|------------:|------------:|------------:|-------------:|
|
217 |
+
| test_loss | 0.00604774 | 0.00604774 | 0.00604774 | 0.00604774 |
|
218 |
+
| test_ANG | 0 | 0 | 0 | 1 |
|
219 |
+
| test_DUC | 0 | 0 | 0 | 3 |
|
220 |
+
| test_FAC | 0.357143 | 0.454545 | 0.4 | 11 |
|
221 |
+
| test_GPE | 0.781726 | 0.789744 | 0.785714 | 195 |
|
222 |
+
| test_LOC | 0.526316 | 0.487805 | 0.506329 | 41 |
|
223 |
+
| test_ORG | 0.785354 | 0.762255 | 0.773632 | 408 |
|
224 |
+
| test_PER | 0.87251 | 0.820225 | 0.84556 | 267 |
|
225 |
+
| test_WOA | 0 | 0 | 0 | 6 |
|
226 |
+
| test_overall_precision | 0.791295 | 0.791295 | 0.791295 | 0.791295 |
|
227 |
+
| test_overall_recall | 0.76073 | 0.76073 | 0.76073 | 0.76073 |
|
228 |
+
| test_overall_f1 | 0.775711 | 0.775711 | 0.775711 | 0.775711 |
|
229 |
+
| test_overall_accuracy | 0.964642 | 0.964642 | 0.964642 | 0.964642 |
|
230 |
+
| test_runtime | 49.5152 | 49.5152 | 49.5152 | 49.5152 |
|
231 |
+
| test_samples_per_second | 23.286 | 23.286 | 23.286 | 23.286 |
|
232 |
+
| test_steps_per_second | 11.653 | 11.653 | 11.653 | 11.653 |
|
233 |
+
| epoch | 2 | 2 | 2 | 2 |
|
234 |
+
|
235 |
### Framework Versions
|
236 |
- Python: 3.10.12
|
237 |
- SpanMarker: 1.5.0
|
|
|
242 |
|
243 |
## Citation
|
244 |
|
245 |
+
|
246 |
```
|
247 |
+
@article{10.1162/tacl_a_00404,
|
248 |
+
author = {Bareket, Dan and Tsarfaty, Reut},
|
249 |
+
title = "{Neural Modeling for Named Entities and Morphology (NEMO2)}",
|
250 |
+
journal = {Transactions of the Association for Computational Linguistics},
|
251 |
+
volume = {9},
|
252 |
+
pages = {909-928},
|
253 |
+
year = {2021},
|
254 |
+
month = {09},
|
255 |
+
abstract = "{Named Entity Recognition (NER) is a fundamental NLP task, commonly formulated as classification over a sequence of tokens. Morphologically rich languages (MRLs) pose a challenge to this basic formulation, as the boundaries of named entities do not necessarily coincide with token boundaries, rather, they respect morphological boundaries. To address NER in MRLs we then need to answer two fundamental questions, namely, what are the basic units to be labeled, and how can these units be detected and classified in realistic settings (i.e., where no gold morphology is available). We empirically investigate these questions on a novel NER benchmark, with parallel token- level and morpheme-level NER annotations, which we develop for Modern Hebrew, a morphologically rich-and-ambiguous language. Our results show that explicitly modeling morphological boundaries leads to improved NER performance, and that a novel hybrid architecture, in which NER precedes and prunes morphological decomposition, greatly outperforms the standard pipeline, where morphological decomposition strictly precedes NER, setting a new performance bar for both Hebrew NER and Hebrew morphological decomposition tasks.}",
|
256 |
+
issn = {2307-387X},
|
257 |
+
doi = {10.1162/tacl_a_00404},
|
258 |
+
url = {https://doi.org/10.1162/tacl\_a\_00404},
|
259 |
+
eprint = {https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl\_a\_00404/1962472/tacl\_a\_00404.pdf},
|
260 |
}
|
261 |
```
|
262 |
|