---
dataset_info:
  features:
  - name: output
    dtype: string
  - name: input
    dtype: string
  splits:
  - name: train
    num_bytes: 343992
    num_examples: 742
  download_size: 172900
  dataset_size: 343992
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
language:
- ar
pretty_name: CATT benchmark
---

CATT automatic diacritization benchmark introduced in https://arxiv.org/abs/2407.03236

Dataset was downloaded from [the CATT repo](https://github.com/abjadai/catt/blob/main/benchmarking/all_models_CATT_data/CATT_data_gt.txt), besides adding a column without diacritics no changes were made.

Refer to the repo for licensing information.