Update README.md
Browse files
README.md
CHANGED
@@ -160,19 +160,19 @@ Note: For French and Italian, the training data remains in German, while the tes
|
|
160 |
|
161 |
### Results
|
162 |
|
163 |
-
Making use of an unsupervised training approach, Swissbert for Sentence Embeddings achieves comparable results as the best-performing multilingual Sentence-BERT model in the semantic textual similarity task for German and outperforms it in the French text classification task.
|
164 |
-
|
165 |
-
| Evaluation task |swissbert | |sentence swissbert| |
|
166 |
-
|
167 |
-
| |accuracy |f1-score |accuracy |f1-score |accuracy
|
168 |
-
| Semantic Similarity DE | 83.80 | - |**93.70** | - | 87.70
|
169 |
-
| Semantic Similarity FR | 82.30 | - |**92.90** | - | 91.10
|
170 |
-
| Semantic Similarity IT | 83.00 | - |**91.20** | - | 89.80
|
171 |
-
| Semantic Similarity RM | 78.80 | - |**90.80** | - | 67.90
|
172 |
-
| Text Classification DE | 95.76 | 91.99 | 96.36 |**92.11**| 95.61
|
173 |
-
| Text Classification FR | 94.55 | 88.52 | 95.76 |**90.94**| 94.55
|
174 |
-
| Text Classification IT | 93.48 | 88.29 | 95.44 | 90.44 | 95.91
|
175 |
-
| Text Classification RM | | | | |
|
176 |
|
177 |
#### Baseline
|
178 |
|
|
|
160 |
|
161 |
### Results
|
162 |
|
163 |
+
Making use of an unsupervised training approach, Swissbert for Sentence Embeddings achieves comparable results as the best-performing multilingual Sentence-BERT model (distiluse-base-multilingual-cased) in the semantic textual similarity task for German and outperforms it in the French text classification task.
|
164 |
+
|
165 |
+
| Evaluation task |swissbert | |sentence swissbert| |Sentence-BERT| |
|
166 |
+
|------------------------|----------|---------|------------------|---------|-------------|---------|
|
167 |
+
| |accuracy |f1-score |accuracy |f1-score |accuracy |f1-score |
|
168 |
+
| Semantic Similarity DE | 83.80 | - |**93.70** | - | 87.70 | - |
|
169 |
+
| Semantic Similarity FR | 82.30 | - |**92.90** | - | 91.10 | - |
|
170 |
+
| Semantic Similarity IT | 83.00 | - |**91.20** | - | 89.80 | - |
|
171 |
+
| Semantic Similarity RM | 78.80 | - |**90.80** | - | 67.90 | - |
|
172 |
+
| Text Classification DE | 95.76 | 91.99 | 96.36 |**92.11**| 95.61 | 91.20 |
|
173 |
+
| Text Classification FR | 94.55 | 88.52 | 95.76 |**90.94**| 94.55 | 89.82 |
|
174 |
+
| Text Classification IT | 93.48 | 88.29 | 95.44 | 90.44 | 95.91 |**92.05**|
|
175 |
+
| Text Classification RM | | | | | | |
|
176 |
|
177 |
#### Baseline
|
178 |
|