---
language:
- tr
- en
- fr
- de
- es
- it
- ar
- ja
- ru
- zh
tags:
- finance
- financial-news
- multilingual
- regression
- xlmr
license: apache-2.0
pipeline_tag: text-classification
model_type: xlm-roberta
---

# fin-impact-xlmr

**fin-impact-xlmr** is a multilingual regression model that estimates the **financial relevance / impact** of news headlines.

The model outputs a **continuous score between 0 and 10**, where higher values indicate stronger financial relevance.

---

## What does this model do?

Given a short text (typically a news headline), the model predicts how strongly the content is related to:

- financial markets
- macroeconomics
- corporate finance

This is **not a sentiment model**.

---

## Supported Languages

- Turkish
- English
- French
- German
- Spanish
- Italian
- Arabic
- Japanese
- Russian
- Chinese

Language tags are **not required** in the input.

---

## Evaluation (approx.)

- **R2**: ~0.62  
- **MAE**: ~1.0 (on 0–10 scale)

---

## Example

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

tokenizer = AutoTokenizer.from_pretrained("stocky-ai/fin-impact-xlmr")
model = AutoModelForSequenceClassification.from_pretrained("stocky-ai/fin-impact-xlmr")

def predict(text):
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
    with torch.no_grad():
        return model(**inputs).logits.item() * 10

print(predict("Central bank raised interest rates amid inflation concerns"))