Naphon/Thai-English-Corpus

---

Thai-English Corpus

Thai-English Corpus is a large-scale bilingual corpus containing Thai and English text collected from publicly available datasets on Hugging Face.

The corpus combines educational content, web documents, Wikipedia articles, legal texts, medical articles, financial documents, software-related text, and other general-domain content into a unified format suitable for language model pretraining and NLP research.

Each document is stored with its original source dataset to preserve attribution and provenance.

Dataset Structure

Features

Column	Type	Description
text	string	Document text
source	string	Original Hugging Face dataset URL

Example

{
  "text": "ประเทศไทยเป็นประเทศในภูมิภาคเอเชียตะวันออกเฉียงใต้...",
  "source": "https://huggingface.co/datasets/wikimedia/wikipedia"
}

Data Sources

The corpus currently includes content from the following datasets:

Dataset
HuggingFaceFW/finepdfs-edu
HuggingFaceFW/fineweb-2 (Thai)
wikimedia/wikipedia (Thai)
wikimedia/wikipedia (English)
wannaphong/KhanomTanLLM-pretrained-dataset-thai-subset
PawinC/nectec_thaifood
ninyawee/thaifin-financials
ThaiLLM/med-articles
iapp/rag_thai_laws
andstor/the_pile_github

Each record retains its original source through the source column.

Processing

The dataset is constructed by:

1. Streaming source datasets directly from Hugging Face.

2. Extracting the primary text field from each dataset.

3. Converting all records into a common schema:

   • text
   • source

4. Removing empty records.

5. Filtering documents shorter than dataset-specific minimum lengths.

6. Exporting the processed records into Parquet shards.

No manual labeling or annotation has been added.

Domain Coverage

The corpus contains text from a variety of domains, including:

• General web content
• Educational materials
• Wikipedia articles
• Legal documents
• Medical articles
• Financial reports
• Software and programming content
• Technical documentation
• Open-source code repositories and related text
• Food and lifestyle content

Intended Uses

This dataset is intended for:

• Foundation model pretraining
• Continued pretraining
• Tokenizer training
• Embedding model training
• Language modeling research
• Retrieval and representation learning research
• General NLP experimentation

Limitations

This corpus is aggregated from multiple independently collected datasets and may contain:

• Duplicate documents
• Near-duplicate documents
• OCR artifacts
• Formatting inconsistencies
• Boilerplate content
• Outdated information
• Inaccurate information
• Biased or unbalanced content
• Source-specific quality issues

The maintainers have not manually reviewed individual documents.

Users should evaluate suitability for their intended use cases.

Licensing and Attribution

This repository aggregates content originating from multiple upstream datasets.

All original licenses, attribution requirements, and usage restrictions remain applicable to their respective source datasets.

The maintainers of Thai-English Corpus do not claim ownership of the original source material.

Users are responsible for reviewing and complying with the licenses and terms of all upstream datasets before using this corpus.

Loading the Dataset

from datasets import load_dataset

dataset = load_dataset("Naphon/Thai-English-Corpus")

Citation

If you use this dataset, please cite both this repository and the original source datasets whenever applicable.

@dataset{thai_english_corpus,
  title={Thai-English Corpus},
  year={2026},
  publisher={Hugging Face}
}