README.md

Dataset Card for HumAID

Table of Contents

• Dataset Description
  • Dataset Summary
  • Supported Tasks
  • 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

Dataset Description

• Homepage: https://crisisnlp.qcri.org/humaid_dataset
• Repository: https://crisisnlp.qcri.org/data/humaid/humaid_data_all.zip
• Paper: https://ojs.aaai.org/index.php/ICWSM/article/view/18116/17919

Dataset Summary

The HumAID Twitter dataset consists of several thousands of manually annotated tweets that has been collected during 19 major natural disaster events including earthquakes, hurricanes, wildfires, and floods, which happened from 2016 to 2019 across different parts of the World. The annotations in the provided datasets consists of following humanitarian categories. The dataset consists only english tweets and it is the largest dataset for crisis informatics so far. ** Humanitarian categories **

• Caution and advice
• Displaced people and evacuations
• Dont know cant judge
• Infrastructure and utility damage
• Injured or dead people
• Missing or found people
• Not humanitarian
• Other relevant information
• Requests or urgent needs
• Rescue volunteering or donation effort
• Sympathy and support

The resulting annotated dataset consists of 11 labels.

Supported Tasks and Benchmark

The dataset can be used to train a model for multiclass tweet classification for disaster response. The benchmark results can be found in https://ojs.aaai.org/index.php/ICWSM/article/view/18116/17919.

Dataset is also released with event-wise and JSON objects for further research. Full set of the dataset can be found in https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/A7NVF7

Languages

English

Dataset Structure

Data Instances

{

"tweet_text": "@RT_com: URGENT: Death toll in #Ecuador #quake rises to 233 \u2013 President #Correa #1 in #Pakistan", 

"class_label": "injured_or_dead_people"

}

Data Fields

• tweet_text: corresponds to the tweet text.
• class_label: corresponds to a label assigned to a given tweet text

Data Splits

• Train
• Development
• Test

Dataset Creation

Source Data

Initial Data Collection and Normalization

Tweets has been collected during several disaster events.

Annotations

Annotation process

AMT has been used to annotate the dataset. Please check the paper for a more detail.

Who are the annotators?

• crowdsourced

Additional Information

Dataset Curators

Authors of the paper.

Licensing Information

• cc-by-nc-4.0

Citation Information

@inproceedings{humaid2020,
Author = {Firoj Alam, Umair Qazi, Muhammad Imran, Ferda Ofli},
booktitle={Proceedings of the Fifteenth International AAAI Conference on Web and Social Media},
series={ICWSM~'21},
Keywords = {Social Media, Crisis Computing, Tweet Text Classification, Disaster Response},
Title = {HumAID: Human-Annotated Disaster Incidents Data from Twitter},
Year = {2021},
publisher={AAAI},
address={Online},
}