title	Coptic Translation Interface
emoji	🔮
colorFrom	green
colorTo	indigo
sdk	docker
app_port	7860
pinned	false
license	apache-2.0
short_description	Coptic↔English translation + neural-symbolic parser

🔮 Coptic Translation & Parsing Interface

A comprehensive research tool for Coptic language analysis combining neural machine translation with neural-symbolic dependency parsing.

Overview

This interface provides three integrated tools for Coptic language research:

1. Neural Machine Translation (Coptic ↔ English)
2. Neural-Symbolic Dependency Parser (Stanza + Prolog)
3. Grammatical Validation (Walter Till's grammar + Crum's lexicon)

Features

🔄 Translation (Coptic ↔ English)

• Coptic → English: megalaa/coptic-english-translator
• English → Coptic: megalaa/english-coptic-translator
• Dialects: Sahidic (literary standard) and Bohairic (liturgical)
• Virtual Keyboard: 31 Coptic Unicode characters
• Example Corpus: Simple sentences, complex structures, and full texts
• Models: Fine-tuned MarianMT on 50,000+ CopticScriptorium parallel sentences

📊 Dependency Parsing (Neural-Symbolic Hybrid)

• Neural Layer:

  • Stanza NLP pipeline for Coptic
  • Tokenization, POS tagging, lemmatization
  • DiaParser for dependency trees

• Symbolic Layer:

  • Prolog implementation of Walter Till's Coptic Grammar (1955)
  • Integration with Crum's Coptic Dictionary (1939)
  • Grammatical pattern detection (tripartite sentences, etc.)
  • Error detection for neural parser hallucinations

• Export: CoNLL-U format for corpus linguistics research

🔍 Grammatical Validation

• Detects grammatical patterns from Walter Till's grammar
• Validates dependency structures against linguistic rules
• Identifies common parsing errors and hallucinations
• Provides grammatical warnings and suggestions

Installation

Prerequisites

• Python 3.10+
• SWI-Prolog 8.0+ (for Prolog validation)
• Docker (for containerized deployment)

Local Installation

# Clone the repository
git clone https://github.com/Rogaton/coptic-translation-interface.git
cd coptic-translation-interface

# Install dependencies
pip install -r requirements.txt

# Download Stanza Coptic models
python -c "import stanza; stanza.download('cop')"

# Run the interface
python app.py

Docker Deployment

# Build the Docker image
docker build -t coptic-interface .

# Run the container
docker run -p 7860:7860 coptic-interface

Usage

Web Interface

Access the interface at http://localhost:7860 with three tabs:

1. Coptic → English: Translate Coptic text to English
2. English → Coptic: Translate English text to Coptic
3. Dependency Analysis: Parse Coptic text with neural-symbolic validation

Example: Translation

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Load model
tokenizer = AutoTokenizer.from_pretrained("megalaa/coptic-english-translator")
model = AutoModelForSeq2SeqLM.from_pretrained("megalaa/coptic-english-translator")

# Translate
coptic_text = "ⲡϫⲟⲉⲓⲥ ⲡⲉ ⲡⲁⲛⲟⲩⲧⲉ"
inputs = tokenizer(coptic_text, return_tensors="pt")
outputs = model.generate(**inputs)
translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
# Output: "The Lord is my God"

Example: Dependency Parsing

from coptic_parser_core import CopticParserCore

# Initialize parser
parser = CopticParserCore()
parser.load_parser()

# Parse text
result = parser.parse_text("ⲁⲛⲟⲕ ⲡⲉ ⲡⲛⲟⲩⲧⲉ")

# Export to CoNLL-U
conllu = parser.format_conllu(result)

Architecture

Neural-Symbolic Hybrid

Input Text
    ↓
[Neural Layer]
    ↓
Stanza Pipeline → Tokenization, POS, Lemmatization
    ↓
DiaParser → Dependency Trees
    ↓
[Symbolic Layer]
    ↓
Prolog Rules → Grammatical Validation
    ↓
Till Grammar + Crum Lexicon → Error Detection
    ↓
Output: Validated Parse + Warnings

Translation Models

Both translation models use:

• Architecture: MarianMT (Seq2Seq Transformer)
• Training Data: CopticScriptorium parallel corpus (50,000+ sentences)
• Preprocessing: Coptic Unicode → Greek transcription
• Dialect Tags: Cyrillic markers (з for Sahidic, б for Bohairic)

Data

Test Corpus

The interface includes coptic_test_corpus.json with:

• Simple Sentences: 10+ examples (Sahidic & Bohairic)
• Complex Sentences: 5+ examples with subordination
• Full Texts: Biblical narratives and parables
• Grammar Patterns: Tripartite nominals, perfect tense, etc.

Linguistic Resources

• Walter Till's Grammar (coptic_grammar.pl): 700+ Prolog rules
• Crum's Dictionary (coptic_lexicon.pl): 12,000+ lexical entries
• Stanza Models: Pre-trained on CopticScriptorium Universal Dependencies

Research Applications

• Corpus Linguistics: CoNLL-U export for quantitative analysis
• Digital Humanities: Automated parsing of Coptic manuscripts
• Language Learning: Interactive translation with grammatical feedback
• Computational Linguistics: Neural-symbolic architecture research
• Egyptology: Analysis of Coptic Biblical and documentary texts

Performance

• Translation Quality: BLEU score ~35-40 (Coptic→English)
• Parsing Accuracy: UAS ~85%, LAS ~80% on CopticScriptorium test set
• Prolog Validation: Detects 70%+ of common parsing errors
• Inference Speed: ~0.5s per sentence (translation), ~2s (parsing with validation)

Citation

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

@software{linden2025coptic,
  author = {Linden, André},
  title = {Coptic Translation and Parsing Interface: A Neural-Symbolic Approach},
  year = {2025},
  publisher = {Zenodo},
  doi = {10.5281/zenodo.19487216},
  url = {https://doi.org/10.5281/zenodo.19487216},
  version = {1.0.1}
}

Related Publications

• Enis, M. & Megalaa, A. (2024). Ancient voices, modern technology: Low-resource neural machine translation for coptic texts. [Paper Link]
• Till, W. C. (1955). Koptische Grammatik (Saïdischer Dialekt). Leipzig: VEB Verlag Enzyklopädie.
• Crum, W. E. (1939). A Coptic Dictionary. Oxford: Clarendon Press.

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Components

• Translation Models: megalaa models (various licenses)
• Stanza: Apache License 2.0
• Prolog Rules: CC BY-NC-SA 4.0 (based on Till's grammar)
• Lexicon: Public domain (Crum's dictionary)

Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/improvement)
3. Commit your changes (git commit -m 'Add new feature')
4. Push to the branch (git push origin feature/improvement)
5. Open a Pull Request

Areas for Contribution

• Additional dialect support (Akhmimic, Lycopolitan, Fayyumic)
• Improved Prolog rules for complex constructions
• Enhanced error detection algorithms
• Extended test corpus with documentary texts
• Performance optimizations

Acknowledgments

• CopticScriptorium for the parallel corpus and UD annotations
• Amir Zeldes and team for Coptic NLP resources
• Stanford NLP Group for Stanza
• megalaa for the translation models
• Walter Till for the foundational Coptic grammar
• W. E. Crum for the comprehensive dictionary

Links

• Live Demo: HuggingFace Space
• Translation Models: megalaa/coptic-english-translator, megalaa/english-coptic-translator
• Dependency Parser: coptic-dependency-parser
• GitHub Profile: @Rogaton

Contact

André Linden

• Email: linden@bluewin.ch
• GitHub: @Rogaton

Version History

• v1.0.1 (2025-04-09): Zenodo archive release
  • DOI: 10.5281/zenodo.19487216
  • Updated documentation with contact information
  • All features from v1.0.0
• v1.0.0 (2025-04-09): Initial release
  • Neural machine translation (Coptic ↔ English)
  • Neural-symbolic dependency parser
  • Prolog grammatical validation
  • Web interface with examples

---

Built with ❤️ for Coptic language research and digital humanities