README.md

---
license: apache-2.0
task_categories:
  - text-classification
language:
  - en
size_categories:
  - 100K<n<1M
---
# SuperQualityDataset
<!-- markdownlint-disable first-line-h1 -->
<!-- markdownlint-disable html -->
<!-- markdownlint-disable no-duplicate-header -->

<div align="center">
  <img src="images/fig1.png" width="60%" alt="SuperQualityDataset" />
</div>
<hr>

<div align="center" style="line-height: 1;">
  <a href="LICENSE" style="margin: 2px;">
    <img alt="License" src="images/fig2.png" style="display: inline-block; vertical-align: middle;"/>
  </a>
</div>

## 1. Introduction

SuperQualityDataset is a comprehensive text classification dataset that has undergone rigorous quality improvements. In the latest version, the dataset features enhanced data cleaning pipelines, improved label accuracy, and better coverage of edge cases. The dataset demonstrates outstanding quality metrics across various evaluation criteria.

<p align="center">
  <img width="80%" src="images/fig3.png">
</p>

Compared to previous versions, this dataset shows significant improvements in data quality. For instance, the label accuracy has increased from 91.2% in v1 to 98.7% in the current version. This improvement comes from enhanced annotation guidelines and multiple rounds of quality verification.

Beyond improved accuracy, this version also offers reduced noise, better class balance, and comprehensive metadata annotations.

## 2. Quality Metrics

### Comprehensive Quality Assessment

<div align="center">

| | Metric | Dataset1 | Dataset2 | Dataset1-v2 | SuperQualityDataset |
|---|---|---|---|---|---|
| **Data Integrity** | Accuracy | 0.912 | 0.925 | 0.931 | 0.940 |
| | Completeness | 0.856 | 0.871 | 0.882 | 0.904 |
| | Consistency | 0.823 | 0.845 | 0.861 | 0.922 |
| **Distribution Quality** | Diversity | 0.745 | 0.762 | 0.778 | 0.892 |
| | Label Distribution | 0.689 | 0.712 | 0.725 | 0.800 |
| | Value Coverage | 0.812 | 0.834 | 0.848 | 0.880 |
| | Temporal Coverage | 0.756 | 0.778 | 0.791 | 0.869 |
| **Content Quality** | Text Quality | 0.834 | 0.856 | 0.867 | 0.925 |
| | Format Validity | 0.901 | 0.912 | 0.923 | 0.960 |
| | Schema Compliance | 0.878 | 0.891 | 0.902 | 0.964 |
| | Uniqueness | 0.823 | 0.845 | 0.856 | 0.893 |
| **Noise Metrics** | Noise Level | 0.789 | 0.812 | 0.823 | 0.872 |
| | Outlier Ratio | 0.756 | 0.778 | 0.789 | 0.888 |
| | Redundancy | 0.801 | 0.823 | 0.834 | 0.857 |
| | Timeliness | 0.867 | 0.878 | 0.889 | 0.938 |

</div>

### Overall Quality Summary
SuperQualityDataset demonstrates exceptional quality across all evaluation dimensions, with particularly strong results in data integrity and content quality metrics.

## 3. Dataset Statistics
- **Total Samples**: 500,000
- **Training Set**: 400,000 samples
- **Validation Set**: 50,000 samples
- **Test Set**: 50,000 samples
- **Number of Classes**: 12
- **Average Text Length**: 156 tokens

## 4. How to Use

```python
from datasets import load_dataset

dataset = load_dataset("SuperQualityDataset")

# Access splits
train_data = dataset["train"]
val_data = dataset["validation"]
test_data = dataset["test"]
```

### Data Format
Each sample contains:
- `text`: The input text string
- `label`: Integer class label (0-11)
- `metadata`: Additional annotation information

### Preprocessing Recommendations
We recommend the following preprocessing steps:
1. Lowercase conversion for case-insensitive tasks
2. Basic tokenization using standard tokenizers
3. Length filtering for samples exceeding 512 tokens

## 5. License
This dataset is released under the Apache 2.0 License. Commercial use and derivative works are permitted with proper attribution.

## 6. Citation
If you use this dataset, please cite:
```
@dataset{superqualitydataset2025,
  title={SuperQualityDataset: A High-Quality Text Classification Dataset},
  author={Quality Team},
  year={2025}
}
```

## 7. Contact
For questions or issues, please open a GitHub issue or email us at contact@superqualitydataset.ai.