mkotan/electra-base-turkish-thquad

ELECTRA-Base Turkish THQuAD – Question Answering Model

This model is a fine-tuned version of dbmdz/electra-base-turkish-cased-discriminator on the Turkish Question Answering task using the THQuAD dataset.
It is designed to answer questions based on a given Turkish context, similar to the original SQuAD format.

🧠 Model Details

• Base Model: dbmdz/electra-base-turkish-cased-discriminator
• Task: Extractive Question Answering (SQuAD-style)
• Dataset: THQuAD (Turkish version of SQuAD)
• Training Epochs: 5
• Batch Size: 12
• Learning Rate: 3e-5
• Max Seq Length: 384
• Framework: Hugging Face Transformers

📌 Citation

Please cite the following resources:

@article{doi:10.5505/pajes.2025.44459,
author = {Sazak, Halenur and Kotan, Muhammed},
title = {Transformer-Based Question Answering Systems for Higher Education: A Comparative Study of Turkish and Multilingual Models},
journal = {Pamukkale Univ Muh Bilim Derg},
volume = { },
number = { },
pages = {0-0},
year = { },
doi = {10.5505/pajes.2025.44459},
note ={doi: 10.5505/pajes.2025.44459},
URL = {https://dx.doi.org/10.5505/pajes.2025.44459},
}

}

🚀 Example Usage

from transformers import pipeline, AutoTokenizer, AutoModelForQuestionAnswering

model_name = "mkotan/electra-base-turkish-thquad"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForQuestionAnswering.from_pretrained(model_name)
qa_pipeline = pipeline("question-answering", model=model, tokenizer=tokenizer)

context = """
Your context here...
"""

print(qa_pipeline(question=" your question here...", context=context))
print(qa_pipeline(question="your question here...", context=context))

✅ Try it Online

You can interact with this model on Hugging Face Spaces:
🔗 Try it here