File size: 1,772 Bytes
6ce6794
bda61e4
 
 
 
 
 
 
 
 
 
 
 
 
 
6ce6794
 
bda61e4
6ce6794
bda61e4
6ce6794
1109a41
9182166
6ce6794
bda61e4
6ce6794
bda61e4
6ce6794
bda61e4
 
 
 
 
6ce6794
bda61e4
6ce6794
bda61e4
6ce6794
bda61e4
6ce6794
bda61e4
 
 
 
 
 
 
 
 
6ce6794
bda61e4
6ce6794
bda61e4
6ce6794
bda61e4
 
6ce6794
961da01
6ce6794
bda61e4
 
6ce6794
1109a41
bda61e4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
---
language: ko
license: mit
tags:
  - sentiment
  - korean
  - finance
  - binary-classification
  - text-classification
datasets:
  - custom
pipeline_tag: text-classification
widget:
  - text: "환율 급등 우려로 시장 불안이 커지고 있다."
  - text: "주가 상승 기대감에 투자심리가 개선되고 있다."
---

# FinBERT Sentiment KO (Binary) - `finbert-sentiment-ko-binary`

🧠 **FinBERT 기반 한국어 감정 분석 이진 분류 모델 (긍정/부정)**

이 모델은 `snunlp/KR-FinBERT-SC`를 기반으로 파인튜닝된 모델로, **환율 뉴스에 대한 긍정/부정 감정 분류**를 수행합니다.  
`중립` 라벨은 제외하고, 오직 긍정(1)과 부정(0) 두 가지 감정만 예측하도록 학습되었습니다.

---

## 🔧 모델 구조

- **Base model**: [`snunlp/KR-FinBERT-SC`](https://huggingface.co/snunlp/KR-FinBERT-SC)
- **Task**: Text Classification (Binary Sentiment)
- **Classes**:
  - `0`: 부정 (Negative)
  - `1`: 긍정 (Positive)

---

## 📊 성능 평가

테스트셋 기준 성능 (42개 샘플):

| Metric         | Score |
|----------------|-------|
| Accuracy       | 0.95  |
| Precision (부정) | 0.92 |
| Recall (부정)    | 1.00 |
| F1-score (부정) | 0.96 |
| Precision (긍정) | 1.00 |
| Recall (긍정)    | 0.89 |
| F1-score (긍정) | 0.94 |

---

## 🧪 사용 예시

```python
from transformers import pipeline

classifier = pipeline("text-classification", model="DataWizardd/finbert-sentiment-binary")

print(classifier("환율 급등 우려로 시장 불안이 커지고 있다."))
# 출력 예시: [{'label': 'LABEL_0', 'score': 0.98}]  ← 부정

print(classifier("환율 급락이 예상된다."))
# 출력 예시: [{'label': 'LABEL_1', 'score': 0.95}]  ← 긍정