Syrian Arabic Dialect LoRA – AMIYA Submission
This repository contains a LoRA adapter for modeling Syrian Arabic dialect,
submitted to the AMIYA (Arabic Modeling In Your Accent) Shared Task.
Base Model
Method
- Prompt-guided dialectal rewriting
- LoRA fine-tuning
- Post-generation filtering to ensure dialect fidelity
- No factual or QA guarantees (dialect fidelity only)
Training Data
- Task-provided data
- Prompt-engineered inference
- No off-limits datasets were used
Intended Use
- Dialectal Arabic generation (Syrian)
- Evaluation under the AL-QASIDA benchmark
Limitations
- This model is not designed for factual question answering.
- Semantic adequacy is out of scope; focus is on dialect fidelity.
Shared Task
- AMIYA – Arabic Modeling In Your Accent
- Dialect: Syrian Arabic
