README.md

---
license: apache-2.0
datasets:
- MIMEDIS/newton-media-unlabeled
language:
- sk
base_model:
- gerulata/slovakbert
pipeline_tag: text-classification
tags:
- Stance
---



# Stance Detection Model for Slovak

This model is fine-tuned from [gerulata/slovakbert](https://huggingface.co/gerulata/slovakbert) for stance detection on Slovak text. It classifies text into three stance categories: Negative, Neutral, and Positive.

## Model Details

- **Base Model**: gerulata/slovakbert
- **Task**: Stance Detection / Sentiment Classification
- **Language**: Slovak (sk)
- **Number of Labels**: 3

## Label Mappings

| Label ID | Stance |
|----------|--------|
| 0 | Negative |
| 1 | Neutral |
| 2 | Positive |

## Usage

### Quick Start

```python
from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch

# Load model and tokenizer
model_name = "MIMEDIS/stance-headlines-model"
model = AutoModelForSequenceClassification.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

# Example text
text = "Toto je skvelý nápad a plne ho podporujem!"

# Tokenize and predict
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
outputs = model(**inputs)
predictions = torch.softmax(outputs.logits, dim=-1)

# Get predicted label
label_id = torch.argmax(predictions, dim=-1).item()
label_map = {0: "Negative", 1: "Neutral", 2: "Positive"}

print(f"Text: {text}")
print(f"Predicted stance: {label_map[label_id]}")
print(f"Confidence scores: Negative={predictions[0][0]:.3f}, Neutral={predictions[0][1]:.3f}, Positive={predictions[0][2]:.3f}")
```