File size: 2,403 Bytes
11cb4d1 c036e3c 11cb4d1 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 | ---
language:
- egy
- de
tags:
- translation
- ancient-egyptian
- hieroglyphics
- contamination-study
- nlp4dh
license: mit
base_model: facebook/mbart-large-50-many-to-many-mmt
pipeline_tag: translation
---
# mBART-50 — Hieroglyphic-to-German Translation
This model is part of the paper **"Data Contamination in Neural Machine Translation of Ancient Egyptian Hieroglyphics"** (NLP4DH 2026).
## Model Description
mBART-50 (611M) fine-tuned for cross-architecture validation (lr=3e-5, AdamW, cosine warmup, 3x A6000 GPUs). Best checkpoint at step 2,000.
**Task:** Ancient Egyptian hieroglyphics (Gardiner notation) → German translation
## Performance
| Subset | BLEU |
|--------|------|
| All (n=50) | 42.5 |
| Contaminated (n=16) | 72.8 |
| Clean (n=34) | 35.5 |
> **Important:** The "All" and "Contaminated" BLEU scores are inflated due to target-side data contamination (32% of test targets appear in training). The **Clean** score represents genuine translation quality on decontaminated samples.
## Usage
```python
from transformers import M2MForConditionalGeneration, M2MTokenizer
model = M2MForConditionalGeneration.from_pretrained("bumblelbee/hiero-mbart50")
tokenizer = M2MTokenizer.from_pretrained("bumblelbee/hiero-mbart50")
# Gardiner notation input (hieroglyphic transliteration)
source = "D36 N35 G17 D21 X1 O34"
tokenizer.src_lang = "ea"
inputs = tokenizer(source, return_tensors="pt")
generated = model.generate(**inputs, forced_bos_token_id=tokenizer.get_lang_id("de"))
output = tokenizer.decode(generated[0], skip_special_tokens=True)
print(output)
```
## Training Data
Fine-tuned on 18,669 ea→de pairs from the Thesaurus Linguae Aegyptiae (TLA), maintained by the Berlin-Brandenburg Academy of Sciences and Humanities.
## Citation
```bibtex
@inproceedings{toutou-etal-2026-data,
title = {Data Contamination in Neural Hieroglyphic Translation:
A Reproducibility Study},
author = {Toutou, Ammar and Harb, Abdelrahman and Basta, Christine},
booktitle = {Proceedings of the 6th International Conference on
Natural Language Processing for Digital Humanities
(NLP4DH 2026)},
year = {2026},
address = {San Diego, USA},
publisher = {Association for Computational Linguistics},
}
```
## Paper Repository
See the full paper, scripts, and results: [GitHub repository](https://github.com/[repository])
|