README.md

metadata

license: apache-2.0
language:
  - en
  - es

Dataset Card for Emoevent

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
  • Source Data
  • Annotations
• Additional Information
  • Licensing Information
  • Citation Information

Dataset Description

• Repository: EmoEvent dataset repository
• Paper: EmoEvent: A Multilingual Emotion Corpus based on different Events
• Leaderboard: Leaderboard for EmoEvent / Spanish version
• Point of Contact: fmplaza@ujaen.es

Dataset Summary

EmoEvent is a multilingual emotion dataset of tweets based on different events that took place in April 2019. Three annotators labeled the tweets following the six Ekman’s basic emotion model (anger, fear, sadness, joy, disgust, surprise) plus the “neutral or other emotions” category. Morevoer, the tweets are annotated as offensive (OFF) or non-offensive (NO).

Supported Tasks and Leaderboards

This dataset is intended for multi-class emotion classification and binary offensive classification.

Competition EmoEvalEs task on emotion detection for Spanish at IberLEF 2021

Languages

• Spanish
• English

Dataset Structure

Data Instances

For each instance, there is a string for the id of the tweet, a string for the emotion class, a string for the offensive class, and a string for the event. See the to explore more examples.

{'id': 'a0c1a858-a9b8-4cb1-8a81-1602736ff5b8',
 'event': 'GameOfThrones',
 'tweet': 'ARYA DE MI VIDA. ERES MAS ÉPICA QUE EL GOL DE INIESTA JODER #JuegodeTronos #VivePoniente',
 'offensive': 'NO',
 'emotion': 'joy',
 }

{'id': '3YCT0L9OMMFP7KWKQSTJRJO0YHUSN2a0c1a858-a9b8-4cb1-8a81-1602736ff5b8',
 'event': 'GameOfThrones',
 'tweet': 'The #NotreDameCathedralFire is indeed sad and people call all offered donations humane acts, but please if you have money to donate, donate to humans and help bring food to their tables and affordable education first. What more humane than that? #HumanityFirst',
 'offensive': 'NO',
 'emotion': 'sadness',
 }

Data Fields

• id: a string to identify the tweet
• event: a string containing the event associated with the tweet
• tweet: a string containing the text of the tweet
• offensive: a string containing the offensive gold label
• emotion: a string containing the emotion gold label

Data Splits

The EmoEvent dataset has 2 subsets: EmoEvent_es (Spanish version) and EmoEvent_en (English version)

Each subset contains 3 splits: train, validation, and test. Below are the statistics subsets.

EmoEvent_es	Number of Instances in Split
Train	5,723
Validation	844
Test	1,656

EmoEvent_en	Number of Instances in Split
Train	5,112
Validation	744
Test	1,447

Dataset Creation

Source Data

Twitter

Who are the annotators?

Amazon Mechanical Turkers

Additional Information

Licensing Information

The EmoEvent dataset is released under the Apache-2.0 License.

Citation Information

@inproceedings{plaza-del-arco-etal-2020-emoevent,
    title = "{{E}mo{E}vent: A Multilingual Emotion Corpus based on different Events}",
    author = "{Plaza-del-Arco}, {Flor Miriam} and Strapparava, Carlo and {Ure{\~n}a-L{\’o}pez}, L. Alfonso and {Mart{\’i}n-Valdivia}, M. Teresa",
    booktitle = "Proceedings of the 12th Language Resources and Evaluation Conference",
    month = may,
    year = "2020",
    address = "Marseille, France", publisher = "European Language Resources Association",
    url = "https://www.aclweb.org/anthology/2020.lrec-1.186", pages = "1492--1498",
    language = "English",
    ISBN = "979-10-95546-34-4"
}