Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,32 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
|
| 6 |
+
Multi-language sentiment classification model developed over the Microsoft (DeBERTa-v3 base model)[https://huggingface.co/microsoft/deberta-v3-base]. In order to train the model the following dataset where used:
|
| 7 |
+
- (tyqiangz/multilingual-sentiments)[tyqiangz/multilingual-sentiments]
|
| 8 |
+
- (cardiffnlp/tweet_sentiment_multilingual)[cardiffnlp/tweet_sentiment_multilingual]
|
| 9 |
+
- (mteb/tweet_sentiment_multilingual)[mteb/tweet_sentiment_multilingual]
|
| 10 |
+
- (Sp1786/multiclass-sentiment-analysis-dataset)[Sp1786/multiclass-sentiment-analysis-dataset]
|
| 11 |
+
- ABSC amazon review
|
| 12 |
+
- SST2
|
| 13 |
+
|
| 14 |
+
Evaluation and comparison with GPT-4o model:
|
| 15 |
+
|
| 16 |
+
Dataset Model F1 Accuracy
|
| 17 |
+
sst2 Our 0.6161 0.9231
|
| 18 |
+
GPT-4 0.6113 0.8605
|
| 19 |
+
sent-eng Our 0.6289 0.6470
|
| 20 |
+
GPT-4 0.4611 0.5870
|
| 21 |
+
sent-twi Our 0.3368 0.3488
|
| 22 |
+
GPT-4 0.5049 0.5385
|
| 23 |
+
mixed Our 0.5644 0.7786
|
| 24 |
+
GPT-4 0.5336 0.6863
|
| 25 |
+
absc-laptop Our 0.5513 0.6682
|
| 26 |
+
GPT-4 0.6679 0.7642
|
| 27 |
+
absc-rest Our 0.6149 0.7726
|
| 28 |
+
GPT-4 0.7057 0.8385
|
| 29 |
+
stanford Our 0.8352 0.8353
|
| 30 |
+
GPT-4 0.8045 0.8032
|
| 31 |
+
amazon-var Our 0.6432 0.9647
|
| 32 |
+
GPT-4 - 0.9450
|