---
library_name: transformers
tags:
- finance
- twitter
- prediction
- ner
- named-entity-recognition
- turkish
- transformers
- bert
license: mit
base_model:
- dbmdz/bert-base-turkish-128k-uncased
---

# 🧠 Prediction Phrase Extractor (NER)

This is a fine-tuned **Named Entity Recognition (NER)** model that extracts **stock prediction phrases** from Turkish financial tweets. These prediction phrases are later passed into a sentiment classifier for further analysis.

🧾 Example predictions:
- `"will reach 70 TL"`  
- `"to moon soon"`  
- `"drop to 50 in 2 weeks"`

---

## 🧠 Model Details

- **Developed by:** damlakonur  
- **Model type:** `BERT` fine-tuned for `token-classification`  
- **Language(s):** Turkish  
- **Finetuned from:** `bert-base-cased`  
- **Entity type:** `Tahmin` (prediction phrase)  
- **License:** MIT  

---

## 🚀 How to Use

```python
from transformers import pipeline

model = pipeline(
    "token-classification",
    model="your-username/prediction-text-ner-bist30",
    aggregation_strategy="simple"
)

model("EREGL will reach 45 TL in June.")