metadata

task_categories:
  - text-generation
language:
  - it
dataset_info:
  features:
    - name: question
      dtype: string
    - name: answer
      dtype: string
  splits:
    - name: train
      num_bytes: 35023539
      num_examples: 17057
  download_size: 19022936
  dataset_size: 35023539
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Sonnet 3.5 🇮🇹 Dataset 📊 Model Card

Overview

The Sonnet 3.5 🇮🇹 Dataset is a high-quality 🏆 dataset for various NLP 🗣️ tasks in 🇮🇹. It includes programming 💻 code, instruction-following 📜 examples, Q&A ❓❗ pairs, and translations 🌍, providing rich data to enhance 🇮🇹 language models.

Key Features

Language: Primarily in 🇮🇹, ideal for 🇮🇹 NLP.
Content Types:
- Code: 💻 snippets with explanations.
- Instruction Following: 📜 Examples for task automation 🤖 and problem-solving 🧠.
- Q&A: ❓❗ Pairs covering various topics.
- Translations: 🌍 Examples to improve multilingual understanding.
- General Examples: Summarization ✍️, classification 🏷️, paraphrasing 🔄, etc.

Dataset Composition

Number of Examples: 75,000 📊 examples, with high-quality 🏆 annotations.
Source Diversity: Data from open-source 📂 repositories, user 👤 interactions, and crowd-sourced 🤝 contributions.

Quality Assurance

Consistency: Uniform formatting 📐 and annotation 📝.
High-Quality Responses: Vetted for linguistic quality 🏅 and correctness ✅.
Ethical Considerations: Harmful content 🚫 removed.

License

Released under CC BY-SA 4.0 📜, encouraging both academic 🎓 and commercial 💼 use.

Dataset Creation Cost

Cost: The creation of this dataset required significant investment, with costs reaching approximately 3,000 euros 💶. This reflects the effort, time, and expertise involved in curating high-quality examples. If you're interested in using this dataset, please reach out via email 📧. Please note, this is a private dataset.

Citation

If you use this dataset, please cite:

@dataset{sonnet_3.5_italian,
  title={Sonnet 3.5 Italian Dataset},
  year={2024},
  publisher={DeepMount00},
  note={https://huggingface.co/DeepMount00}
}