open-compass-OpenFinData

<div align="center">

# OpenFinData

大语言模型开源金融评测数据集
</div>

## 项目介绍
OpenFinData是由东方财富与上海人工智能实验室联合发布的开源金融评测数据集。该数据集代表了最真实的产业场景需求，是目前场景最全、专业性最深的金融评测数据集。它基于东方财富实际金融业务的多样化丰富场景，旨在为金融科技领域的研究者和开发者提供一个高质量的数据资源。

## 核心优势

- **真实性**: 数据集专注于金融领域知识，每一条数据均由实际金融业务场景产生，确保了数据的真实性和实用性。
- **全面性**: 涵盖多种数据类型和多样化金融场景，详细归类，确保评估的全面性，满足不同金融模型的需求。
- **专业性**: 数据集由金融行业专业人士构建，经过高质量筛选和处理，保证了数据的专业性和准确性。
- **拓展性**: 数据集设计具有高度的可拓展性，支持快速编辑与迭代，以适应金融科技领域的快速发展和变化。

## 数据集内容

OpenFinData数据集由以下六个模块构成，每个模块包含多个任务维度，以满足金融领域的不同评测需求。以下是数据集内容的详细表格：

| 模块名称 | 任务维度 | 描述 |
|----------------|--------------------------------|--------------------------------------------------------------|
| 金融知识 | 金融术语/黑话 | 提供金融行业中的专业术语及其解释。 |
| | 金融事实 | 记录金融市场的历史事件和事实数据。 |
| 金融判别 | 金融意图理解 | 识别用户在金融场景中的意图。 |
| | 情绪识别 | 分析金融文本中的情绪倾向。 |
| | 金融实体识别 | 从文本中识别出金融相关的实体。 |
| | 金融实体消歧 | 解决金融实体在不同上下文中的歧义问题。 |
| 金融计算 | 金融数据检查 | 验证金融数据的一致性和准确性。 |
| | 金融数值提取 | 从非结构化数据中提取关键的金融数值信息。 |
| | 金融指标计算 | 计算并提供各种金融指标。 |
| 金融分析 | 股票分析 | 对股票市场进行深入分析。 |
| | 基金分析 | 分析基金的表现和风险。 |
| | 行业/板块分析 | 对特定行业或板块的市场表现进行评估。 |
| | 行情分析 | 提供市场行情的综合分析。 |
| 金融解读 | 宏观解读 | 解读宏观经济政策和事件对金融市场的影响。 |
| | 行业解读 | 分析特定行业的发展动态和趋势。 |
| | 公告解读 | 对上市公司公告进行内容分析和影响评估。 |
| | 事件解读 | 解释金融市场中的重大事件及其潜在影响。 |
| 金融合规 | 金融业务合规 | 确保金融业务流程符合相关法规要求。 |
| | 信息安全合规 | 保护金融数据的安全，防止数据泄露和滥用。 |

## 使用指南

1. **数据下载**: 请点击[数据集链接](https://github.com/open-compass/OpenFinData/releases/download/release/openfindata_release.zip)下载数据集。
2. **模型评估**: 我们提供基于[OpenCompass](https://github.com/open-compass/opencompass)的模型评估方案，更多细节将会更新在OpenCompass的文档中。

## 致谢

我们感谢所有参与OpenFinData项目的贡献者，包括但不限于数据收集、清洗、标注和评测的团队成员。特别感谢东方财富和上海人工智能实验室的支持。

## 备注

本数据集仅供学术研究使用，请勿将此数据集用于任何的模型训练。

## 联系方式

如有任何问题或建议，请通过以下方式联系我们：

- Email: [[e-mail]](mailto:[e-mail])
- GitHub Issues: [OpenFinData GitHub页面](https://github.com/open-compass/OpenFinData/issues)

## 更新日志

- **2023年12月29日**: 发布了OpenFinData数据集的初始版本，包含1500条数据记录。

## 下载方法
:modelscope-code[]{type="sdk"}
:modelscope-code[]{type="git"}

<div align="center">
# OpenFinData
Open-Source Financial Evaluation Dataset for Large Language Models
</div>

## Project Introduction
OpenFinData is an open-source financial evaluation dataset co-released by East Money and Shanghai AI Laboratory. This dataset represents the most realistic industrial scenario demands, and is currently the most comprehensive in scenarios and most professional among existing financial evaluation datasets. Built upon the diverse and rich scenarios of East Money's actual financial business, it aims to provide high-quality data resources for researchers and developers in the fintech field.

## Core Advantages
- **Authenticity**: The dataset focuses on financial domain knowledge, with every record generated from actual financial business scenarios, ensuring its authenticity and practicality.
- **Comprehensiveness**: It covers multiple data types and diverse financial scenarios, with detailed classification, ensuring comprehensive evaluation and meeting the requirements of different financial models.
- **Professionalism**: Constructed by financial industry professionals, the dataset has undergone high-quality screening and processing, guaranteeing its professionalism and accuracy.
- **Scalability**: The dataset is designed with high scalability, supporting rapid editing and iteration to adapt to the rapid development and changes in the fintech field.

## Dataset Content
The OpenFinData dataset consists of the following six modules, each containing multiple task dimensions to meet different evaluation needs in the financial field. The detailed table of the dataset content is as follows:

| Module Name | Task Dimension | Description |
|---------------------------|-----------------------------------------|-----------------------------------------------------------------------------|
| Financial Knowledge | Financial Terms/Slang | Provide professional terms and their explanations in the financial industry.|
| | Financial Facts | Record historical events and factual data of financial markets. |
| Financial Discrimination | Financial Intention Understanding | Identify user intentions in financial scenarios. |
| | Sentiment Recognition | Analyze the emotional tendency in financial texts. |
| | Financial Entity Recognition | Identify finance-related entities from texts. |
| | Financial Entity Disambiguation | Resolve ambiguity issues of financial entities in different contexts. |
| Financial Calculation | Financial Data Verification | Verify the consistency and accuracy of financial data. |
| | Financial Value Extraction | Extract key financial value information from unstructured data. |
| | Financial Indicator Calculation | Calculate and provide various financial indicators. |
| Financial Analysis | Stock Analysis | Conduct in-depth analysis of the stock market. |
| | Fund Analysis | Analyze the performance and risks of funds. |
| | Industry/ Sector Analysis | Evaluate the market performance of specific industries or sectors. |
| | Market Quotation Analysis | Provide comprehensive analysis of market quotations. |
| Financial Interpretation | Macroeconomic Interpretation | Interpret the impact of macroeconomic policies and events on financial markets.|
| | Industry Interpretation | Analyze the development dynamics and trends of specific industries. |
| | Announcement Interpretation | Conduct content analysis and impact assessment of listed company announcements.|
| | Event Interpretation | Explain major events in financial markets and their potential impacts. |
| Financial Compliance | Financial Business Compliance | Ensure that financial business processes comply with relevant regulatory requirements.|
| | Information Security Compliance | Protect the security of financial data, prevent data leakage and abuse. |

## Usage Guidelines
1. **Data Download**: Please click [Dataset Link](https://github.com/open-compass/OpenFinData/releases/download/release/openfindata_release.zip) to download the dataset.
2. **Model Evaluation**: We provide a model evaluation solution based on [OpenCompass](https://github.com/open-compass/opencompass), and more details will be updated in the documentation of OpenCompass.

## Acknowledgements
We thank all contributors to the OpenFinData project, including but not limited to team members involved in data collection, cleaning, annotation and evaluation. Special thanks to East Money and Shanghai AI Laboratory for their support.

## Note
This dataset is for academic research use only. Please do not use this dataset for any model training.

## Contact Information
If you have any questions or suggestions, please contact us through the following channels:
- Email: [[e-mail]](mailto:[e-mail])
- GitHub Issues: [OpenFinData GitHub Page](https://github.com/open-compass/OpenFinData/issues)

## Changelog
- **December 29, 2023**: Released the initial version of the OpenFinData dataset, containing 1500 data records.

## Download Methods
:modelscope-code[]{type="sdk"}
:modelscope-code[]{type="git"}

搜集汇总

背景与挑战

背景概述

OpenFinData是一个由东方财富和上海人工智能实验室联合发布的开源金融评估数据集，专注于金融领域知识，数据源自真实金融业务场景，确保真实性和实用性。该数据集覆盖金融知识、判断、计算、分析、解读和合规六大模块，包含多个任务维度如术语解释、情感分析和股票分析等，旨在全面评估大型语言模型在金融场景下的性能，但仅限学术研究使用，不可用于模型训练。