Mistake-To-Meaning

# Clear Spelling Dataset

## Overview

The **Mistake to Meaning** (M2M) dataset is a carefully crafted synthetic collection of **100,000 unique English spelling mistakes and their correct forms**, intended for training high-quality typo correction and spell checking AI models. It covers various types of common mistakes observed frequently in real-world scenarios, such as:

- Keyboard adjacency typos
- Letter swaps and omissions
- Duplicate characters
- Phonetic substitution errors
- Commonly confused homophones (e.g., "their" vs. "there")

## Dataset Format

The dataset is provided in **CSV format** with two clearly defined columns:

| Column | Description | Example |
|----------|---------------------------------------------|---------------------|
| `error` | The misspelled or incorrect word or phrase | "teh" |
| `correct`| The correct word or intended phrase | "the" |

## Usage

This dataset is ideal for:

- Training and fine-tuning **typo correction** models
- Benchmarking **spell-checking algorithms**
- Enhancing NLP model robustness to real-world noisy input

## Quality Assurance

- **No duplicates:** Each (error, correct) pair is unique.
- **Hand-curated seed set:** Includes hundreds of common misspellings verified against real-world usage patterns.
- **Realistic noise generation:** Uses realistic error transformations mimicking genuine human typing behavior.

## License (MIT)

This dataset is released under the permissive **MIT License**, which allows commercial and non-commercial use, distribution, and modification. Attribution is required:

## Citation

If you use this dataset in your research or projects, please provide attribution similar to:

```
This [your project type] uses the Mistake to Learning dataset by ProCreations.
```

Enjoy training your typo-correction models!

# 清晰拼写数据集

## 概述
**错词转语义（Mistake to Meaning，简称M2M）** 数据集是一套精心打造的合成数据集，包含10万组独特的英语拼写错误及其对应正确形式，旨在用于训练高质量的拼写错误纠正与拼写检查人工智能模型。该数据集覆盖现实场景中高频出现的各类常见错误类型，具体包括：
- 键盘相邻按键误输
- 字母调换与遗漏
- 字符重复
- 语音替换错误
- 易混淆同音异义词（例如"their"与"there"）

## 数据集格式
本数据集采用**逗号分隔值（CSV）**格式提供，包含两个明确定义的字段：

| 字段名 | 描述 | 示例 |
|----------|--------------------------------------|---------------------|
| `error` | 拼写错误或不规范的单词/短语 | "teh" |
| `correct`| 正确的单词或预期表达的短语 | "the" |

## 使用场景
本数据集适用于以下场景：
- 训练与微调**拼写错误纠正**模型
- 对**拼写检查算法**开展基准测试
- 提升自然语言处理（Natural Language Processing，简称NLP）模型对现实噪声输入的鲁棒性

## 质量保障
- **无重复条目**：每一组（错误项，正确项）对均唯一。
- **人工精选种子集（hand-curated seed set）**：包含数百组经现实使用模式验证的常见拼写错误。
- **逼真噪声生成**：采用模拟真实人类打字行为的合理错误转换规则。

## 许可证（MIT）
本数据集采用宽松的**MIT许可证**发布，允许商业与非商业使用、分发及修改，但需注明来源：

## 引用说明
若您在研究或项目中使用本数据集，请采用如下格式进行标注：

本[您的项目类型]使用了由ProCreations发布的Mistake to Learning数据集。

祝您的拼写纠错模型训练顺利！