This dataset contains the test split for AlexandriaX Subtask 3, a dialectal Arabic MT evaluation task.
Participants receive machine-translated outputs with source information. The goal is to detect and classify translation errors using LQM-inspired annotations.
Each row contains:
{
"id": "43571_1",
"direction": "ENG_UAE",
"source": "Come on, be with her. It is nonsense. You do… See the full description on the dataset page:
https://huggingface.co/datasets/UBC-NLP/AlexandriaX_Subtask_3_Test.
