FurqonAryadana commited on
Commit
59e9b1a
·
verified ·
1 Parent(s): c046f04

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +61 -0
README.md ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - multilabel-classification
4
+ - emotion-detection
5
+ - text-classification
6
+ - transformers
7
+ - deberta
8
+ - huggingface
9
+ license: apache-2.0
10
+ datasets:
11
+ - custom
12
+ language:
13
+ - en
14
+ ---
15
+
16
+ # DeBERTa-v3-Large for Multilabel Emotion Classification
17
+
18
+ This model is a fine-tuned version of [`microsoft/deberta-v3-large`](https://huggingface.co/microsoft/deberta-v3-large) for multilabel emotion classification.
19
+ It was trained on a custom dataset where each text sample may contain multiple emotions.
20
+
21
+ ## 📌 Model Card
22
+
23
+ - **Model**: `FurqonAryadana/deberta-emotion-multilabel-0.5007`
24
+ - **Base**: DeBERTa-v3-Large
25
+ - **Task**: Multilabel Emotion Classification
26
+ - **License**: Apache 2.0
27
+ - **Language**: English
28
+ - **Threshold Tuning**: Applied per label
29
+
30
+ ## 📊 Evaluation (Validation Set)
31
+
32
+ **Detailed Classification Report (After Threshold Tuning)**:
33
+
34
+ | Emotion | Precision | Recall | F1-score | Support |
35
+ |----------------|-----------|--------|----------|---------|
36
+ | amusement | 0.62 | 0.70 | 0.66 | 851 |
37
+ | anger | 0.41 | 0.61 | 0.49 | 999 |
38
+ | annoyance | 0.33 | 0.77 | 0.46 | 1609 |
39
+ | caring | 0.44 | 0.55 | 0.49 | 594 |
40
+ | confusion | 0.46 | 0.71 | 0.56 | 800 |
41
+ | disappointment | 0.30 | 0.50 | 0.37 | 990 |
42
+ | disgust | 0.28 | 0.46 | 0.35 | 584 |
43
+ | embarrassment | 0.38 | 0.31 | 0.34 | 308 |
44
+ | excitement | 0.42 | 0.52 | 0.47 | 632 |
45
+ | fear | 0.51 | 0.52 | 0.51 | 321 |
46
+ | gratitude | 0.82 | 0.74 | 0.78 | 955 |
47
+ | joy | 0.45 | 0.58 | 0.50 | 876 |
48
+ | love | 0.66 | 0.78 | 0.71 | 701 |
49
+ | sadness | 0.44 | 0.51 | 0.47 | 714 |
50
+
51
+ - **F1 Micro**: `0.5104`
52
+ - **F1 Macro**: `0.5125`
53
+
54
+ ## 🧠 Emotions (Label Order)
55
+
56
+ ```python
57
+ [
58
+ 'amusement', 'anger', 'annoyance', 'caring', 'confusion', 'disappointment',
59
+ 'disgust', 'embarrassment', 'excitement', 'fear', 'gratitude', 'joy',
60
+ 'love', 'sadness'
61
+ ]