北落师门情感数据集version1

# 情感冲击力评估数据集
[](https://opensource.org/licenses/Apache-2.0)

## 数据集概述
本数据集用于训练和评估文本内容对典型核心受众的情感冲击力分析模型，聚焦**内容驱动的社会心理影响评估**。基于评价理论、语用学原则与社会建构主义视角，通过认知评价和语境推理识别文本的潜在情绪影响。

## 数据集结构
```json
{
"instruction": "分析给定文本对其典型核心受众的情感冲击力，输出包含情感冲击力标签列表和对应强度(1-7)列表的JSON字符串。",
"input": "Layin n bed with a headache ughhhh...waitin on your call...",
"output": "{\"labels\": [\"悲伤\", \"无奈\", \"焦虑\"], \"intensities\": [4, 3, 4]}"
}
```

## 标签体系
### 情感标签（20类）
| 类别 | 典型标签 |
|---------------------|---------------------------------|
| 基础情绪 | 快乐、悲伤、愤怒、恐惧、惊讶、厌恶 |
| 复杂认知情绪 | 焦虑、失望、希望、信任、满足、羞耻 |
| 社会性情绪 | 同情、尴尬、幸灾乐祸、轻蔑、自豪 |
| 文化相关情绪 | 委屈、无奈、纠结、心疼 |

### 强度评分（1-7分制）
| 分值 | 描述 | 示例场景 |
|------|---------------------|-------------------------|
| 1 | 微弱/几乎无感 | "今天多云" |
| 4 | 中等影响 | "会议延期" |
| 7 | 极端影响 | "地震致千人伤亡" |

## 典型应用场景
```json
// 社交媒体分析
{"input": "Funeral ceremony...gloomy friday...", "output": {"labels":["悲伤"], "intensities":[5]}}

// 用户意图理解
{"input": "wants to hang out with friends SOON!", "output": {"labels":["快乐","希望"], "intensities":[4,4]}}
```

## 重要原则
1. **内容优先规则**：
✅ 中性公告"股东减持" → 焦虑(5)
❌ 积极措辞"敬请谅解" → 不直接标记积极

2. **多标签标注**：允许"悲伤+无奈"等组合

3. **强度动态校准**：
`航班取消` → 商务旅客(6) vs 游客(4)

## 许可与联系
**License**: [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0)
**Copyright**: Arain119
**Contact**: [e-mail]

```text
除非另有说明，本数据集默认遵循以下条款：
- 允许商业使用，需保留版权声明
- 允许修改，但修改版本需明确标注
- 作者不对数据质量作任何担保
```

# Emotional Impact Assessment Dataset
[](https://opensource.org/licenses/Apache-2.0)

## Dataset Overview
This dataset is designed for training and evaluating models that analyze the emotional impact of textual content on its typical core audience, focusing on **content-driven socio-psychological impact assessment**. Grounded in appraisal theory, pragmatic principles, and the social constructivist perspective, it identifies the potential emotional effects of texts through cognitive evaluation and contextual reasoning.

## Dataset Structure
json
{
"instruction": "Analyze the emotional impact of the given text on its typical core audience, and output a JSON string containing a list of emotional impact labels and their corresponding intensity scores (1-7).",
"input": "Layin n bed with a headache ughhhh...waitin on your call...",
"output": "{"labels": ["Sadness", "Helplessness", "Anxiety"], "intensities": [4, 3, 4]}"
}

## Label System
### Emotional Labels (20 Categories)
| Category | Typical Labels |
|------------------------|--------------------------------------------------------------------------------|
| Basic Emotions | Joy, Sadness, Anger, Fear, Surprise, Disgust |
| Complex Cognitive Emotions | Anxiety, Disappointment, Hope, Trust, Contentment, Shame |
| Social Emotions | Sympathy, Embarrassment, Schadenfreude, Contempt, Pride |
| Culture-related Emotions | Grievance, Helplessness, Conflict, Heartache |

### Intensity Scoring (1-7 Scale)
| Score | Description | Example Scenario |
|-------|------------------------------|----------------------------------------------------------------------------------|
| 1 | Faint / Barely Perceptible | "It's cloudy today" |
| 4 | Moderate Impact | "Meeting postponed" |
| 7 | Extreme Impact | "Earthquake causes thousands of casualties" |

## Typical Application Scenarios
json
// Social Media Analysis
{"input": "Funeral ceremony...gloomy friday...", "output": {"labels":["Sadness"], "intensities":[5]}}

// User Intent Understanding
{"input": "wants to hang out with friends SOON!", "output": {"labels":["Joy", "Hope"], "intensities":[4,4]}}

## Core Principles
1. **Content-First Rule**
✅ Neutral announcement "Shareholder reduction of holdings" → Anxiety (5)
❌ Positive phrasing "Please understand" → No positive label directly assigned

2. **Multi-Label Annotation**
Combinations such as "Sadness + Helplessness" are allowed.

3. **Intensity Dynamic Calibration**
`Flight cancellation` → Business travelers (6) vs. Tourists (4)

## License and Contact
**License**: [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0)
**Copyright**: Arain119
**Contact**: [e-mail]

text
Unless otherwise specified, this dataset follows the following terms by default:
- Commercial use is permitted, provided that the copyright notice is retained
- Modification is permitted, provided that modified versions are clearly marked
- The author makes no warranties regarding the quality of the dataset

搜集汇总

背景与挑战

背景概述

北落师门情感数据集version1是一个用于多标签情感冲击识别任务的结构化监督数据集，基于DeepSeek-V3教师模型生成，包含来自中英文评论、新闻和股票文本的37万余条样本。每条数据采用指令-输入-输出三元组格式，输出为字符串化JSON，涵盖20类固定情感标签及1-7的强度分值，适用于指令微调和知识蒸馏等应用。