GPT-2 Fine-tuned on Luther Translation Letters

Model Description

This model is a fine-tuned version of GPT-2 trained on Martin Luther's translation letters and related correspondence. The model has been specifically adapted to generate text in the style and language patterns found in Luther's writings about biblical translation, theological discourse, and 16th-century German religious texts.

Model Details

Base Model: GPT-2
Model Size: 81.9M parameters
Fine-tuning Domain: Martin Luther's translation letters and correspondence
Language: Primarily German (16th century style) with some Latin phrases
License: BSD

Intended Use

Primary Use Cases

Historical text generation in Luther's style
Academic research on 16th-century German religious literature
Educational applications for studying Reformation-era language
Stylistic analysis of Luther's translation methodology

Limitations

The model is trained on historical texts and may not be suitable for modern German text generation
Output may contain archaic language forms and spellings
Should not be used as a source of theological or historical facts without verification
May reflect the historical biases and perspectives of the training material

Training Details

Training Data

The model was fine-tuned on a curated dataset of Martin Luther's letters related to his Bible translation work, including:

Correspondence about translation methodology
Letters discussing theological interpretation
Communications with fellow reformers and scholars

Training Procedure

Base model: GPT-2
Fine-tuning approach: Continued pre-training on domain-specific corpus
Framework: Transformers library

Usage

Installation

pip install transformers torch

Basic Usage

from transformers import GPT2LMHeadModel, GPT2Tokenizer

# Load model and tokenizer
model = GPT2LMHeadModel.from_pretrained("ajsbsd/gpt2-finetuned-luther-translation-letter")
tokenizer = GPT2Tokenizer.from_pretrained("ajsbsd/gpt2-finetuned-luther-translation-letter")

# Generate text
input_text = "Lieber Freund, bezüglich der Übersetzung"
input_ids = tokenizer.encode(input_text, return_tensors='pt')

# Generate
with torch.no_grad():
    output = model.generate(
        input_ids,
        max_length=100,
        num_return_sequences=1,
        temperature=0.8,
        do_sample=True,
        pad_token_id=tokenizer.eos_token_id
    )

generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
print(generated_text)

Evaluation

The model has been evaluated on:

Perplexity on held-out Luther correspondence
Stylistic similarity to authentic Luther texts
Coherence of generated theological discourse

Ethical Considerations

Historical Context

This model reproduces language and perspectives from the 16th century
Generated content should be understood within its historical context
Users should be aware of potential anachronistic viewpoints

Responsible Use

Intended for academic, educational, and research purposes
Should not be used to generate content that could be mistaken for authentic historical documents
Users should clearly label generated content as AI-produced

Citation

If you use this model in your research, please cite:

@misc{gpt2-luther-translation,
  author = {ajsbsd},
  title = {GPT-2 Fine-tuned on Luther Translation Letters},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/ajsbsd/gpt2-finetuned-luther-translation-letter}
}

Contact

For questions or issues related to this model, please open an issue on the Hugging Face model page or contact the model author.

Acknowledgments

Original GPT-2 model by OpenAI
Historical texts from Martin Luther's correspondence
Hugging Face Transformers library

Note: This model is provided for educational and research purposes. Generated content should always be verified against historical sources and understood within its proper historical context.

ajsbsd
/

gpt2-finetuned-luther-translation-letter