joker122322222/natural-disasters-from-social-media

---
task_categories:
- text-classification
language:
- en
tags:
- natural disasters
- tweets
- classification
- catastrophic events
pretty_name: Natural Disasters from Social Media
size_categories:
- 100K<n<1M
annotations_creators:
- crowdsourced
- expert-generated
source_datasets:
- "Kaggle 1 - jannesklaas/disasters-on-social-media"
- "Kaggle 2 - vstepanenko/disaster-tweets"
- "Kaggle 3 - sidharth178/disaster-response-messages"
- "Zahra et al. - doi: 10.1016/j.ipm.2019.102107"
- "CrisisMMD - arxiv: 1805.00713"
- "Alam et al. - arxiv: 1805.05151"
- "CrisisLexT26 - doi: 10.1145/2675133.2675242"
- "Imran et al. - aclanthology: L16-1259"
- "CrisisLexT6 - doi: 10.1609/icwsm.v8i1.14538"
- "HumAID - doi: 10.1609/icwsm.v15i1.18116"
- "CrisisBench - doi: 10.1609/icwsm.v15i1.18115"
configs:
- config_name: default
default: true
data_files:
- split: train
path: "train.csv"
- split: validation
path: "validation.csv"
- split: test
path: "test.csv"
- config_name: full
data_files: "meta/natural-disasters-from-social-media.csv"
- config_name: meta
data_files: "meta/distributions/*.csv"

dataset_info:
config_name: default
splits:
- name: train
num_bytes: 39817704
num_examples: 169109
- name: validation
num_bytes: 4977163
num_examples: 21139
- name: test
num_bytes: 4981112
num_examples: 21139
dataset_size: 49775824
---

# Description

Dataset created for Master's thesis "Detection of Catastrophic Events from Social Media" at the Slovak Technical University Faculty of Informatics.

Contains posts from social media that are split into two categories:
- Informative - related and informative in regards to natural disasters
- Non-Informative - unrelated to natural disasters

Other metadata include event type, source dataset etc. To balance classes, 50k tweets from twitter archive for years 2017-2022 were added.

# Distributions

# Source Datasets:

| **Name** | **Count** |
|:----------------------------------------------------------------------------------------:|:---------:|
| Kaggle 1 - [URL](https://www.kaggle.com/datasets/jannesklaas/disasters-on-social-media) | 951 |
| Kaggle 2 - [URL](https://www.kaggle.com/datasets/vstepanenko/disaster-tweets) | 579 |
| Kaggle 3 - [URL](https://www.kaggle.com/datasets/sidharth178/disaster-response-messages) | 3782 |
| Zahra et al. - [URL](https://doi.org/10.1016/j.ipm.2019.102107) | 6494 |
| CrisisMMD - [URL](https://arxiv.org/abs/1805.00713) | 11043 |
| Alam et al. - [URL](https://arxiv.org/abs/1805.05151) | 11133 |
| CrisisLexT26 - [URL](https://doi.org/10.1145/2675133.2675242) | 14998 |
| Imran et al. - [URL](https://aclanthology.org/L16-1259) | 16549 |
| CrisisLexT6 - [URL](https://doi.org/10.1609/icwsm.v8i1.14538) | 22672 |
| HumAID - [URL](https://doi.org/10.1609/icwsm.v15i1.18116) | 42837 |
| CrisisBench - [URL](https://doi.org/10.1609/icwsm.v15i1.18115) | 31158 |
| ArchiveTeam - [URL](https://archive.org/details/twitterstream) | 49191 |
| **Total** | 211387 |

# Total Event counts:

| **Type** | **Non-Informative** | **Informative** | **Total** |
|:----------:|:-------------------:|:---------------:|:---------:|
| Unknown | 61880 | 14740 | 76620 |
| Storm | 20944 | 47301 | 68245 |
| Flood | 13104 | 14637 | 27741 |
| Earthquake | 7844 | 15549 | 23393 |
| Fire | 2343 | 8595 | 10938 |
| Landslide | 2392 | 384 | 2776 |
| Meteorite | 193 | 545 | 738 |
| Haze | 51 | 503 | 554 |
| Volcano | 243 | 139 | 382 |