README.md

metadata

language:
  - fr
license: cc-by-sa-4.0
task_categories:
  - text-classification
dataset_info:
  features:
    - name: previous_sentence
      dtype: string
    - name: types
      sequence: string
    - name: modes
      sequence: string
    - name: categories
      sequence: string
    - name: next_sentence
      dtype: string
    - name: target_sentence
      dtype: string
    - name: is_emotional
      dtype: bool
  splits:
    - name: train
      num_bytes: 6845736
      num_examples: 19560
    - name: validation
      num_bytes: 958060
      num_examples: 2781
    - name: test
      num_bytes: 1969946
      num_examples: 5570
  download_size: 5791557
  dataset_size: 9773742
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
tags:
  - emotions

Dataset Card for [Dataset Name]

Table of Contents

• Table of Contents
• Dataset Description
• Dataset Summary
• Supported Tasks and Leaderboards
• Languages
• Dataset Structure
• Data Instances
• Data Fields
• Data Splits
• Dataset Creation
• Curation Rationale
• Source Data
• Annotations
• Personal and Sensitive Information
• Considerations for Using the Data
• Social Impact of Dataset
• Discussion of Biases
• Other Known Limitations
• Additional Information
• Dataset Curators
• Licensing Information
• Citation Information
• Contributions

Dataset Description

• Homepage:
• Repository:
• Paper: https://arxiv.org/abs/2405.14385
• Leaderboard:
• Point of Contact: Gwénolé Lecorvé

Dataset Summary

EmoTextToKids provides sentences from written documents annotated in emotions. Emotions are characterized by their emotional category (fear, anger, pride...) and their expression mode (labeled, behavioral, displayed or suggester). As opposed to usual datasets in emotion recognition, the documents are not conversational. They are newspapers, encyclopedias, novels, dedicated to children.

Supported Tasks and Leaderboards

• Emotion recognition

Languages

• French

Dataset Structure

Data Instances

{
  "previous_sentence": "Un an plus tard, le Sénat lui accorde la dictature sans limite dans le temps. ",
  "target_sentence": "Mais à Rome, la gloire de César inquiète certains sénateurs. ",
  "next_sentence": "Un complot commence à s’organiser autour d’un homme nommé Cassius. ",
  "is_emotional": true,
  "modes": [
      "labeled"
    ],
  "types": [
      "basic"
    ],
  "categories": [
      "fear"
    ]
}

The fields modes, types and categories are lists because several emotions can be present in a unique sentence.

Data Fields

[More Information Needed]

Data Splits

Subset	Texts	Sent.	Tokens	Emotional sent.
train	1,129	19,553	360K	3,952
dev	182	2,770	53K	438
test	283	5,588	102K	984
Total	1,594	27,911	515K	5,374

Dataset Creation

Curation Rationale

[More Information Needed]

Source Data

Initial Data Collection and Normalization

[More Information Needed]

Who are the source language producers?

[More Information Needed]

Annotations

Annotation process

Data were manually annotated by 6 experts following annotation guidelines here: https://hal.science/hal-03263194 .

Annotations were validated by comparing a significant sample of the annotated data with annotation of an external expert. Below are the kappa coefficients.

Label	Kappa
emotional	0.66
Modes
behavioral	0.70
labeled	0.73
displayed	0.68
suggested	0.46
Types
basic	0.66
complex	0.55
Categories
admiration	0.53
anger	0.71
guilt	0.50
disgust	0.87
embarrassment	0.51
pride	0.25
jealousy	1.00
joy	0.51
fear	0.64

Who are the annotators?

[More Information Needed]

Personal and Sensitive Information

[More Information Needed]

Considerations for Using the Data

Social Impact of Dataset

[More Information Needed]

Discussion of Biases

[More Information Needed]

Other Known Limitations

[More Information Needed]

Additional Information

Dataset Curators

[More Information Needed]

Licensing Information

[More Information Needed]

Citation Information

@misc

@inproceedings{etienne2024emotion,
  title={Emotion Identification for French in Written Texts: Considering Modes of Emotion Expression as a Step Towards Text Complexity Analysis},
  author={{\'E}tienne, Aline and Battistelli, Delphine and Lecorv{\'e}, Gw{\'e}nol{\'e}},
  booktitle={Proceedings of the 14th ACL Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis (WASSA)},
  year={2024}
}