Burmese Dataset Translator

A command-line tool for translating English text datasets into Burmese using the Gemini API.

This tool is designed for AI/ML dataset preparation and performs safe batch translation with automatic error handling and API limit protection.

Features

• Batch translation to reduce API requests
• Token-aware batching for reliable processing
• Automatic retry for temporary API errors
• Safe stop when API quota is exceeded
• Saves translated and untranslated rows separately
• Clean command-line output
• Works with CSV datasets

Installation

You can run the tool without cloning the repository.

Install Globally

Install the CLI globally using npm:

npm install -g dataset-translator

After installation, run the tool from anywhere:

dataset-translator

Run Without Installing (npx)

You can also run the tool directly using npx:

npx dataset-translator

Requirements

• Node.js 18 or later
• Gemini API key

The translator uses the model:

• Gemini 2.5 Flash

Dataset Format

Your CSV file must contain the following columns:

| Column | Description | | ------ | ----------------------------- | | text | English sentence to translate | | label | Classification label |

Example dataset:

text,label
I love programming.,2
This is very sad.,0
The movie was amazing.,1

Usage

Run the command:

dataset-translator

The CLI will ask for:

1. CSV dataset path
2. Gemini API key

Example:

Burmese Dataset Translator

Enter CSV dataset path: example_dataset.csv
Enter Gemini API Key: **********************

Output

The tool creates translated files in batches.

Example output:

burmese_dataset_0-457.csv
burmese_dataset_457-855.csv
untranslated_dataset_855-953.csv

File Types

Translated files

burmese_dataset_*.csv

Contain translated Burmese text.

Untranslated files

untranslated_dataset_*.csv

Contain rows that could not be translated due to API limits.

Error Handling

The translator includes automatic protection against API issues.

Temporary API Errors

The tool retries automatically.

API Quota Exceeded

If the Gemini API quota is reached:

• Translation stops safely
• Remaining rows are saved to untranslated_dataset
• No data is lost

Typical Workflow

1. Prepare dataset CSV file
2. Run the translator
3. Wait for batch translation
4. Use translated dataset for training or analysis

Use Cases

This tool is useful for:

• Sentiment analysis datasets
• NLP dataset localization
• Multilingual AI training
• Dataset augmentation

License

MIT License