DataScouter 1.1.0

An AI-powered dataset search engine across Hugging Face, Kaggle, and Google Dataset Search

DataScouter

DataScouter is a Python library for searching datasets across Hugging Face, Kaggle, and Google Dataset Search using semantic similarity and fuzzy matching.

Features

• Multi-Source Search: Fetch datasets from Hugging Face, Kaggle, and Google Dataset Search.
• Semantic Search: Leverages NLP embeddings for improved relevance.
• Fuzzy Matching: Enhances search results using string similarity techniques.
• Optimized Performance: Utilizes PyTorch for efficient similarity calculations.
• API Key Support: Supports authentication for Kaggle dataset access.

Installation

Install DataScouter via pip:

pip install datascouter

Ensure you have the required dependencies:

pip install requests beautifulsoup4 transformers fuzzywuzzy torch numpy

Usage

Basic Search

from datascouter import DataScouter

# Initialize DataScouter
search_engine = DataScouter(kaggle_api_key="your_kaggle_api_key")

# Search for datasets related to 'climate change'
results = search_engine.search_datasets("climate change")

# Print results
for dataset in results:
    print(f"{dataset['source']}: {dataset['name']} - {dataset['description']} (Score: {dataset['score']})")

Using Environment Variable for Kaggle API Key

Instead of passing the API key directly, set it as an environment variable:

export KAGGLE_API_KEY="your_kaggle_api_key"

Then initialize DataScouter without the API key:

search_engine = DataScouter()

How It Works

1. Fetches dataset metadata from Hugging Face, Kaggle, and Google Dataset Search.
2. Uses sentence-transformers to generate embeddings for semantic similarity.
3. Applies fuzzy matching to refine search results.
4. Filters and ranks datasets based on relevance threshold.

API Reference

DataScouter

Initialization

DataScouter(kaggle_api_key=None, relevance_threshold=0.3, source_filter=None)

• kaggle_api_key (str, optional): API key for accessing Kaggle datasets.
• relevance_threshold (float, default=0.3): Minimum similarity score required for results.
• source_filter (str, optional): Filter results by a specific source ("Hugging Face", "Kaggle", "Google Dataset Search").

search_datasets

search_datasets(query: str) -> list

• query (str): Search term for datasets.
• Returns: A list of matching datasets sorted by relevance.

Contributing

We welcome contributions! To contribute:

1. Fork the repository.
2. Create a feature branch: git checkout -b feature-xyz.
3. Commit changes: git commit -m 'Added new feature'.
4. Push to your branch: git push origin feature-xyz.
5. Submit a Pull Request.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Support

If you find DataScouter useful, consider starring the repository on GitHub. For issues, create a GitHub issue or contact us at [e-mail].