metadata

dataset_info:
  features:
    - name: id
      dtype: int64
    - name: q
      dtype: string
    - name: a
      dtype: string
  splits:
    - name: th2en
      num_bytes: 261404758
      num_examples: 438008
    - name: en2th
      num_bytes: 261020179
      num_examples: 437191
    - name: mtinstruct_th2en
      num_bytes: 194991586
      num_examples: 438008
    - name: mtinstruct_en2th
      num_bytes: 194712580
      num_examples: 437191
  download_size: 349631371
  dataset_size: 912129103
configs:
  - config_name: default
    data_files:
      - split: th2en
        path: data/th2en-*
      - split: en2th
        path: data/en2th-*
      - split: mtinstruct_th2en
        path: data/mtinstruct_th2en-*
      - split: mtinstruct_en2th
        path: data/mtinstruct_en2th-*
license: cc-by-3.0
language:
  - th
  - en

Thai Aligninstruct Dataset

This project was try to make English-Thai Aligninstruct.

I use simalign for making the High Quality Word Alignments Without Parallel Training Data Using Static and Contextualized Embeddings and use deepcut (by lekcut) to tokenize Thai word.

The data use scb-mt-en-th-2020.

References

Aligninstruct: https://arxiv.org/abs/2401.05811v1
SimAlign: High Quality Word Alignments Without Parallel Training Data Using Static and Contextualized Embeddings: https://aclanthology.org/2020.findings-emnlp.147/
scb-mt-en-th-2020: A Large English-Thai Parallel Corpus: https://arxiv.org/abs/2007.03541