jppgks/twitter-financial-news-sentiment
收藏资源简介:
该数据集是为LLM(大型语言模型)微调准备的,包含Twitter上的金融新闻情感分析数据。数据集通过添加一个`instruction`列并将标签从数字映射到字符串(`{0:negative, 1:positive, 2:neutral}`)来进行处理。数据集包含训练集和验证集,分别有9543和2388个样本。每个样本包含`input`(输入文本)、`output`(情感标签)和`instruction`(指令)三个特征。
This dataset is designed for the fine-tuning of Large Language Models (LLMs), and comprises financial news sentiment analysis data sourced from Twitter. It is preprocessed by adding an `instruction` column and converting the original numerical labels into string labels following the mapping `{0: negative, 1: positive, 2: neutral}`. The dataset is split into a training set and a validation set, containing 9543 and 2388 samples respectively. Each sample includes three features: `input` (the input text), `output` (the sentiment label), and `instruction` (the task instruction).
数据集概述
许可证
- MIT许可证
数据集信息
- 特征:
input:字符串类型output:字符串类型instruction:字符串类型
- 数据分割:
train:- 字节数:1906560
- 样本数:9543
validation:- 字节数:479540
- 样本数:2388
- 下载大小:728648字节
- 数据集大小:2386100字节
配置
- 默认配置:
- 数据文件:
train:data/train-*validation:data/validation-*
- 数据文件:




