TimKoornstra/synthetic-financial-tweets-sentiment

---
license: mit
dataset_info:
features:
- name: tweet
dtype: string
- name: sentiment
dtype:
class_label:
names:
'0': neutral
'1': bullish
'2': bearish
splits:
- name: train
num_bytes: 85168816
num_examples: 831530
download_size: 47315785
dataset_size: 85168816
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
task_categories:
- text-classification
language:
- en
tags:
- synthetic
- tweet
- tweets
- sentiment
- classification
- mixtral
- FinTwitBERT
- FinTwitBERT-sentiment
pretty_name: Synthetic Financial Tweets with Sentiment
size_categories:
- 100K<n<1M
---

# FinTwitBERT: Synthetic Financial Tweets Dataset

## Description
This dataset contains a collection of synthetically generated tweets related to financial markets, including discussions on stocks and cryptocurrencies. The tweets were generated using the `NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO` model, employing 10-shot random examples from the `TimKoornstra/financial-tweets-sentiment` dataset. Each entry in this dataset provides insights into financial discussions and is labeled with a sentiment value, aimed at supporting sentiment analysis in the financial domain.

## Dataset Structure
Each record in the dataset is structured as follows:
- **Tweet**: The text of the tweet, offering insights into discussions surrounding financial markets.
- **Sentiment**: A numerical label indicating the sentiment of the tweet, with '1' for bullish, '2' for bearish, and '0' for neutral sentiments.

## Dataset Size
The dataset comprises 1,428,771 tweets in total, categorized into:
- 486,133 bearish tweets
- 486,366 bullish tweets
- 456,272 neutral tweets

## Preprocessing
The dataset has undergone thorough preprocessing to ensure the quality and consistency of the data. This includes sentiment mapping to the numerical labels and the removal of duplicate entries.

## Disclaimer
Given the synthetic nature of this dataset, users should be aware that it may contain shocking, incorrect, or bizarre content, including tweets that do not accurately match their sentiment labels. This dataset is generated through a machine learning model and as such, reflects the model's limitations in understanding complex human sentiments and contexts. Users are advised to apply caution and perform additional validations when using this dataset for research or applications.

## Usage
This dataset is particularly suited for training and evaluating machine learning models focused on sentiment analysis within the financial sector. It can serve as a valuable resource for:
- Academic research in financial sentiment analysis
- Financial market trend analysis
- Development of AI tools for financial institutions

The structured sentiment labels make it ideal for supervised learning approaches aiming to understand market trends and investor sentiment.

## License
This dataset is made available under the MIT License.

## Citation
If you use this dataset in your research or applications, please cite it as follows:

```
@misc{FinTwitBERT,
author = {Stephan Akkerman, Tim Koornstra},
title = {FinTwitBERT: A Specialized Language Model for Financial Tweets},
year = {2024},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/TimKoornstra/FinTwitBERT}}
}
```

Additionally, if you utilize the sentiment classifier trained on this dataset, please cite:

```
@misc{FinTwitBERT-sentiment,
author = {Stephan Akkerman, Tim Koornstra},
title = {FinTwitBERT-sentiment: A Sentiment Classifier for Financial Tweets},
year = {2024},
publisher = {Hugging Face},
howpublished = {\url{https://huggingface.co/StephanAkkerman/FinTwitBERT-sentiment}}
}
```

## Contributions
We welcome contributions to the dataset, including suggestions for improvements, reporting issues, and additional data. Please feel free to reach out or submit pull requests to the dataset repository.

原始信息汇总

FinTwitBERT: Synthetic Financial Tweets Dataset

描述

该数据集包含一系列与金融市场相关的合成推文，涉及股票和加密货币的讨论。这些推文使用 NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO 模型生成，采用 TimKoornstra/financial-tweets-sentiment 数据集中的10次随机示例。每条推文都带有情感标签，旨在支持金融领域的情感分析。

数据集结构

数据集中的每条记录结构如下：

• Tweet: 推文文本，提供有关金融市场讨论的见解。
• Sentiment: 一个数值标签，表示推文的情感，其中 1 表示看涨，2 表示看跌，0 表示中性。

数据集大小

数据集总共包含1,428,771条推文，分类如下：

• 486,133条看跌推文
• 486,366条看涨推文
• 456,272条中性推文

预处理

数据集经过彻底的预处理，以确保数据的质量和一致性。这包括情感映射到数值标签和删除重复条目。

免责声明

鉴于该数据集的合成性质，用户应注意可能包含令人震惊、不正确或怪异的内容，包括与情感标签不准确匹配的推文。该数据集是通过机器学习模型生成的，因此反映了模型在理解复杂人类情感和上下文方面的局限性。用户在使用此数据集进行研究或应用时应谨慎，并进行额外的验证。

使用

该数据集特别适用于训练和评估专注于金融领域情感分析的机器学习模型。它可以作为以下方面的宝贵资源：

• 金融情感分析的学术研究
• 金融市场趋势分析
• 金融机构AI工具的开发

结构化的情感标签使其非常适合采用监督学习方法，旨在理解市场趋势和投资者情绪。

许可证

该数据集在MIT许可证下提供。