Par-Four-Fineweb-Edu-Fortified-Finance

Subset of https://huggingface.co/datasets/Josephgflowers/Par-Four-Fineweb-Edu-Fortified

Used keyword filtering and scoring to extract data with a finance focus.

Creation script:

https://huggingface.co/datasets/Josephgflowers/Par-Four-Fineweb-Edu-Fortified-Finance/resolve/main/find-fin-fine.py


## Dataset Description

### Summary
This dataset is a finance-focused filtered subset of the Fineweb-Edu-Fortified dataset. It is designed to extract and prioritize high-quality educational content relevant to finance, economics, and related topics. Using advanced keyword-based scoring, the dataset emphasizes finance-specific entries while reducing the overall dataset size for more targeted training.

The dataset includes three key fields:
- **score**: Represents the quality and relevance of the text content.
- **text**: The main content of the webpage, filtered for finance-related keywords.
- **url**: The source URL from which the text was extracted.

### Source and Reference
The original dataset, Fineweb-Edu-Fortified, was created from 95 Common Crawl datasets, covering web content from 2013 to 2024. This filtered version maintains its core educational purpose while refining the focus to finance-related materials.

The original dataset card is available [here](https://huggingface.co/datasets/airtrain-ai/fineweb-edu-fortified).

## Supported Tasks and Use Cases

1. **Financial Model Fine-Tuning**
- Train smaller language models for finance-specific applications such as market analysis, economic research, and financial QA.

2. **Synthetic Dataset Creation**
- Generate question-answer pairs and financial problem-solving datasets using extracted content.

3. **Topic-Based Training**
- Focus on specialized finance-related training by grouping the dataset by subject.

4. **Model Healing**
- Fine-tune pruned models to restore or enhance knowledge in financial and economic domains.

## Dataset Structure

### Data Fields
- **score** (int): Quality score representing relevance to finance and economics.
- **text** (string): The main content of the webpage.
- **url** (string): The source URL.

### Format
The dataset is pre-filtered and scored using regex-based financial keywords. It retains only entries meeting a financial relevance threshold.

## Languages
The dataset is predominantly in **English**, but some entries may include multilingual content due to the source data.

## License
This dataset is released under the [Open Data Commons Attribution License (ODC-By) v1.0](https://opendatacommons.org/licenses/by/1.0/), in accordance with the original dataset’s licensing.

### Help Here
Like my work? Want to see more? Custom request? Message me on discord: joseph.flowers.ra Donate here: https://buymeacoffee.com/josephgflowers

## Citation
If you use this dataset, please cite the original Fineweb-Edu-Fortified dataset:

```bibtex
@dataset{airtrain2024finewebedu,
title={Fineweb-Edu-Fortified},
author={Airtrain AI},
year={2024},
url={https://huggingface.co/datasets/airtrain-ai/fineweb-edu-fortified}
}

本数据集为 https://huggingface.co/datasets/Josephgflowers/Par-Four-Fineweb-Edu-Fortified 的子集。
采用关键词过滤与评分机制，提取聚焦金融领域的数据。

创建脚本：
https://huggingface.co/datasets/Josephgflowers/Par-Four-Fineweb-Edu-Fortified-Finance/resolve/main/find-fin-fine.py


## 数据集说明

### 摘要
本数据集是Fineweb-Edu-Fortified数据集的金融领域过滤子集，旨在提取并优先收录与金融、经济及相关主题相关的高质量教育类内容。通过先进的基于关键词的评分机制，该数据集重点聚焦金融专属条目，同时压缩整体数据集规模，以支持更精准的模型训练。

本数据集包含三个核心字段：
- **score（评分）**：表征文本内容的质量与相关性。
- **text（文本）**：网页的主体内容，已针对金融相关关键词完成过滤。
- **url（来源链接）**：提取文本的原始来源URL。

### 来源与参考
原始数据集Fineweb-Edu-Fortified基于95个通用爬取数据集（Common Crawl）构建，涵盖2013年至2024年的网络内容。本过滤版本保留了其核心的教育属性，同时将聚焦范围细化至金融相关材料。
原始数据集卡片可参见[此处](https://huggingface.co/datasets/airtrain-ai/fineweb-edu-fortified)。

## 支持任务与应用场景

1. **金融模型微调**
- 针对金融特定应用训练轻量化大语言模型，例如市场分析、经济研究与金融问答。
2. **合成数据集构建**
- 利用提取的内容生成问答对与金融问题求解数据集。
3. **主题式训练**
- 通过按主题分组数据集，聚焦于专业金融相关的模型训练。
4. **模型修复**
- 对剪枝后的模型进行微调，以恢复或增强其在金融与经济领域的知识储备。

## 数据集结构

### 数据字段
- **score** (int)：表征文本与金融、经济主题相关性的质量评分，数据类型为整数。
- **text** (string)：网页主体内容，数据类型为字符串。
- **url** (string)：原始来源链接，数据类型为字符串。

### 数据格式
本数据集通过基于正则表达式的金融关键词完成过滤与评分，仅保留达到金融相关性阈值的条目。

## 语言分布
本数据集主要以**英语**为主，但由于源数据的特性，部分条目可能包含多语言内容。

## 许可证
本数据集遵循[开放数据通用署名许可证（Open Data Commons Attribution License, ODC-By）v1.0](https://opendatacommons.org/licenses/by/1.0/)，与原始数据集的许可协议保持一致。

### 相关支持
喜欢我的工作？希望获取更多内容？或有定制需求？请在Discord上联系我：joseph.flowers.ra。捐赠通道：https://buymeacoffee.com/josephgflowers

## 引用说明
若您使用本数据集，请引用原始Fineweb-Edu-Fortified数据集：
bibtex
@dataset{airtrain2024finewebedu,
title={Fineweb-Edu-Fortified},
author={Airtrain AI},
year={2024},
url={https://huggingface.co/datasets/airtrain-ai/fineweb-edu-fortified}
}