---
task_categories:
- translation
- text-classification
language:
- zh
---
# Cantonese-Mandarin-Traditional Chinese Parallel Corpus

This dataset provides a parallel corpus of Cantonese, Simplified Chinese, and Traditional Chinese text.

## Dataset Composition

The dataset is a combination of two existing datasets:

1. [botisan-ai/cantonese-mandarin-translations](https://huggingface.co/datasets/botisan-ai/cantonese-mandarin-translations)
2. [raptorkwok/cantonese-chinese-dataset-gen2](https://huggingface.co/datasets/raptorkwok/cantonese-chinese-dataset-gen2)

**Train Set:** Merged from both source datasets

**Test and Validation Sets:** Derived from raptorkwok/cantonese-chinese-dataset-gen2

## Language Variants

The dataset contains three language variants:

1. **yue:** Spoken Cantonese (Yue) rendered into words
2. **zh:** Simplified Chinese
3. **zht:** Traditional Chinese as written in Hong Kong

## Conversion Process

The conversion from Simplified Chinese to Traditional Chinese was performed using [StarCC](https://github.com/StarCC0/spec).

## Data Format

The dataset is structured in JSON format, with each entry containing parallel text in the three language variants. Example entry:

```json
{
  "yue": "講唔到重點而且唔夠全面",
  "zh": "说不了重点而且不够全面",
  "zht": "說不了重點而且不夠全面"
}
```

## Use Cases

This dataset can be valuable for various natural language processing tasks, including:

- Machine translation between Cantonese, Simplified Chinese, and Traditional Chinese
- Comparative linguistic studies of Chinese language variants
- Development of multilingual Chinese language models

## Limitations

Users should be aware that automatic conversion between Simplified and Traditional Chinese, while generally reliable, may not always capture differences in vocabulary and idioms between regions using different writing systems.