FrenchEmotionalNarratives

# French Emotional Narratives

## Introduction

French Emotional Narratives is a French-language corpus of emotional narratives collected for emotion regulation training and annotated by the writers themselves with a discrete emotion plus four psychologically motivated components: *Behavior*, *Feeling*, *Thinking*, and *Reason*.

The dataset is linked to the paper [Emotion Recognition based on Psychological Components in Guided Narratives for Emotion Regulation](https://arxiv.org/abs/2305.10446), which introduced the corpus and studied the impact of components on emotion classification.

## Dataset structure

### Features

Each example has five string fields:

* `Behavior` — observable actions and context (who did what/where/when).
* `Feeling` — subjective and physiological feelings during the event.
* `Thinking` — thoughts during the event.
* `Reason` — needs and values perceived as validated or attacked.
* `Emotion` — discrete label among *colère, peur, tristesse, joie*.

### Example

```json
{
"Behavior": "En cours, un camarade m'interrompt à voix haute.",
"Feeling": "Cœur qui bat vite, mains moites.",
"Thinking": "Je pense qu'il me manque de respect.",
"Reason": "Mon besoin de respect est attaqué.",
"Emotion": "Colère"
}
```

## Data collection

Narratives were written by participants in emotion regulation trainings following the "Cognitive Analysis of Emotion" method described in [this book](https://www.dunod.com/sciences-humaines-et-sociales/manuel-d-analyse-cognitive-emotions-theorie-et-applications) (France, 2005–2022). Each participant describes an emotional event and fills separate answers for Behavior, Feeling, Thinking, and Reason, then labels the primary emotion.

### Anonymization

We replace personal/organizational names, dates, and locations with specific placeholers (e.g., \[PER\] for a person entity) using [CATIE-AQ/NERmemberta-4entities](https://huggingface.co/CATIE-AQ/NERmemberta-4entities), a French NER model.

### Who are the writers?

Primarily master’s students, doctoral students, and teachers (approx. 20–50 years old) studying/working in France.

## Citation

If you use this dataset, please cite:

Gustave Cortal, Alain Finkel, Patrick Paroubek, and Lina Ye. 2023. [Emotion Recognition based on Psychological Components in Guided Narratives for Emotion Regulation](https://arxiv.org/abs/2305.10446). In *Proceedings of the 7th Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature*, pages 72–81, Dubrovnik, Croatia. Association for Computational Linguistics.

# 法语情感叙事数据集（French Emotional Narratives）

## 简介

法语情感叙事数据集（French Emotional Narratives）是一款专为情感调节训练采集的法语情感叙事语料库，由叙事撰写者自行标注，标注内容包含离散情绪标签与四个基于心理学动机的维度：行为（Behavior）、感受（Feeling）、思维（Thinking）与动因（Reason）。

本数据集关联于论文《基于引导式叙事中心理维度的情感识别与情感调节》（Emotion Recognition based on Psychological Components in Guided Narratives for Emotion Regulation，https://arxiv.org/abs/2305.10446），该论文介绍了此语料库，并探究了各维度对情感分类任务的影响。

## 数据集结构

### 特征字段

每条样本包含五个字符串类型字段：
* `Behavior` — 可观测的行为与情境（即何人、何事、何地、何时发生）。
* `Feeling` — 事件发生过程中的主观感受与生理体验。
* `Thinking` — 事件发生过程中的所思所想。
* `Reason` — 被感知为受到肯定或侵犯的需求与价值观。
* `Emotion` — 选自以下离散情绪标签的主情绪：愤怒（colère）、恐惧（peur）、悲伤（tristesse）、喜悦（joie）。

### 示例

json
{
"Behavior": "En cours, un camarade m'interrompt à voix haute.",
"Feeling": "Cœur qui bat vite, mains moites.",
"Thinking": "Je pense qu'il me manque de respect.",
"Reason": "Mon besoin de respect est attaqué.",
"Emotion": "Colère"
}

## 数据采集

叙事内容由情感调节训练的参与者撰写，遵循2005年至2022年法国出版的《情感认知分析》（Cognitive Analysis of Emotion）一书中所阐述的方法，相关书籍链接为https://www.dunod.com/sciences-humaines-et-sociales/manuel-d-analyse-cognitive-emotions-theorie-et-applications。每位参与者需描述一则情感事件，并分别填写行为、感受、思维与动因四个维度的内容，最后标注出主导情绪。

### 匿名化处理

我们使用法语命名实体识别（Named Entity Recognition，简称NER）模型[CATIE-AQ/NERmemberta-4entities](https://huggingface.co/CATIE-AQ/NERmemberta-4entities)，将个人/组织名称、日期与地理位置替换为特定占位符（例如，用`[PER]`表示人物实体）。

### 撰写者群体

撰写者主要为在法国学习或工作的硕士研究生、博士研究生与教师，年龄大致在20至50岁之间。

## 引用规范

若您使用本数据集，请引用以下文献：

Gustave Cortal、Alain Finkel、Patrick Paroubek与Lina Ye，2023年。《基于引导式叙事中心理维度的情感识别与情感调节》（Emotion Recognition based on Psychological Components in Guided Narratives for Emotion Regulation，https://arxiv.org/abs/2305.10446）。见：《第七届SIGHUM联合研讨会论文集：面向文化遗产、社会科学、人文与文学的计算语言学》（Proceedings of the 7th Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature），克罗地亚杜布罗夫尼克，第72-81页，计算语言学协会（Association for Computational Linguistics）。