README.md

---
annotations_creators:
- expert-generated
language_creators:
- found
multilinguality:
- monolingual
license:
- odc-by
task_categories:
- token-classification
language:
- en
tags:
- climate
pretty_name: Climate Change NER
size_categories:
- n<1K
---
# Dataset Card for Climate Change NER

The Climate Change NER is an English-language dataset containing 534 abstracts of climate-related papers. They have been sourced from the Semantic Scholar Academic Graph "abstracts" dataset. The abstracts have been manually annotated by classifying climate-related tokens in a set of 13 categories.

## Dataset Details

### Dataset Description

We introduce a comprehensive dataset for developing and evaluating NLP models tailored towards understanding and addressing climate-related topics across various domains.
The **Climate Change NER** is a manually curated dataset of 534 abstracts from climate-related articles. They have been extracted from the [Semantic Scholar Academic Graph](https://www.semanticscholar.org/product/api) (_abstracts_ dataset) 
by using a seed set of climate-related keywords. The abstracts were annotated by tagging relevant tokens with the IOB format (inside, outside, beginning) using 14 differents categories:

- **Curated by:** Birgit Pfitzmann
- **Shared by:** Cesar Berrospi Ramis
- **Language(s) (NLP):** English
- **License:** [Open Data Commons Attribution License (ODC-By)](https://opendatacommons.org/licenses/by/1-0/)

### Dataset Sources

- **Source:** A subset of [Semantic Scholar Academic Graph Dataset](https://arxiv.org/abs/2301.10140) (the "abstracts" dataset)
- **Paper:** [INDUS: Effective and Efficient Language Models for Scientific Applications](https://arxiv.org/abs/2405.10725)

## Uses

This dataset is primarly intended for evaluating Named Entity Recognition (NER) tasks.

## Dataset Structure

### Data Instances

Each instance consists of several lines representing a document abstract and its annotation. The beginning of an instance is marked by a line starting with `-DOCSTART-`
and containing a unique hash of the document across the entire dataset. Each token from the abstract is included in a separate line with one of the following labels:
- `O` if the token is not part of any named entity instance
- `B` if the token is the beginning of a named entity instance
- `I` if the token is inside of a named entity instance

If the token is marked with `O` or `I`, the entity type is added. A maximum of one entity type can be assigned to a token. The possible types are:
`climate-assets`, `climate-datasets`, `climate-greenhouse-gases`, `climate-hazards`, `climate-impacts`, `climate-mitigations`, `climate-models`, `climate-nature`, `climate-observations`, `climate-organisms`, `climate-organizations`, `climate-problem-origins`, and `climate-properties`.

```
-DOCSTART- -X- O O cc560f5c553e1b60e054abe1578227ff

Non O
- O
identifiable O
RHH O
emergency O
department O
data O
and O
climate O
data O
from O
the O
Australian B-climate-organizations
Bureau I-climate-organizations
of I-climate-organizations
Meteorology I-climate-organizations
were O
obtained O
for O
the O
period O
2003 O
- O
2010 O
. O
```

### Data Splits

The Climate Change NER dataset has 3 splits: _train_, _validation_, and _test_. Below are the statistics for Version 1.0.0 of the dataset.

| Dataset Split | Number of Instances in Split |
| :------------ | :--------------------------- |
| Train         | 382                          |
| Validation    |  77                          |
| Test          |  75                          |


## Dataset Creation

### Source Data

The source data are abstracts from the [Semantic Scholar Academic Graph Dataset](https://www.semanticscholar.org/product/api) (the _abstracts_ dataset).

### Annotations

The abstracts have been annotated manually. Climate-related tokens are classified according to the following classes:

- **climate-assets**: objects or services of value to humans that can get destroyed or diminished by climate-hazards. Key categories are health, buildings, infrastructure, and crops or livestock.
- **climate-datasets**: specific collections of climate data with a name. A climate dataset can be the result of observations or of a model, e.g., as a prediction or reanalysis. The data may be lists, tables, databases, inventories or historical records, where the data dominate over attached code.
- **climate-greenhouse-gases**: gases that cause heating of the atmosphere (greenhouse gases).
- **climate-hazards**: hazards with potential negative impact on climate, such as floods, wildfires, droughts, and heatwaves. Where a hazard is named in more detail in a text, the entire term is annotated, e.g., _surface water flood_ or _soil liquefaction_.
- **climate-impacts**: effects of hazards, primarily negative effects on humans. We also consider impacts on livestock as impacts, as it indirectly affects humans.
- **climate-mitigations**: activities to reduce climate change or to better deal with the consequences. 
- **climate-models**: specific physical, mathematical, or artificial intelligence objects, nowadays always computer-executable, used to analyze and usually predict climate parameters.
- **climate-nature**: aspects of nature that are not alive, such as oceans, rivers, the atmosphere, winds, and snow. 
- **climate-observations**: climate observation tools with a name. Examples are satellites, radiospectrometers, rain gauges, wildlife cameras, and questionnaires.
- **climate-organisms**: animals, plants, and other organisms that are considered for their own sakes (in contrast to as food for humans) as climate organisms.
- **climate-organizations**: real-world organizations with climate-related interests.
- **climate-problem-origins**: problems that describe why the climate is changing. Key examples are _fossil fuel_ and _deforestation_. We also mention sectors that can be cited as causes of energy use. For instance, in a text about the energy consumption by the transport sector,
_transport sector_ is annotated as _problem_.
- **climate-properties**: properties of the climate itself (not abstract objects like models and datasets) that typically come with values and units.

#### Who are the annotators?

The dataset was annotated by Birgit Pfitzmann

## Citation

**BibTeX:**

When using this data, please cite the following paper:

```
@misc{bhattacharjee2024indus,
  title={INDUS: Effective and Efficient Language Models for Scientific Applications}, 
  author={Bishwaranjan Bhattacharjee and Aashka Trivedi and Masayasu Muraoka and Muthukumaran Ramasubramanian and Takuma Udagawa and Iksha Gurung and Rong Zhang and Bharath Dandala and Rahul Ramachandran and Manil Maskey and Kayleen Bugbee and Mike Little and Elizabeth Fancher and Lauren Sanders and Sylvain Costes and Sergi Blanco-Cuaresma and Kelly Lockhart and Thomas Allen and Felix Grazes and Megan Ansdel and Alberto Accomazzi and Yousef El-Kurdi and Davis Wertheimer and Birgit Pfitzmann and Cesar Berrospi Ramis and Michele Dolfi and Rafael Teixeira de Lima and Panos Vagenas and S. Karthik Mukkavilli and Peter Staar and Sanaz Vahidinia and Ryan McGranaghan and Armin Mehrabian and Tsendgar Lee},
  year={2024},
  eprint={2405.10725},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2405.10725}
}
```

The source data has been fetched from the Semantic Scholar Academic Graph Datasets ("abstracts" dataset).
This collection is licensed under [ODC-BY](https://opendatacommons.org/licenses/by/1.0/).
By downloading this data you acknowledge that you have read and agreed to all the terms in this license.
When using this data in a product or service, or including data in a redistribution, please cite the following paper:

```
@misc{https://doi.org/10.48550/arxiv.2301.10140,
  title = {The Semantic Scholar Open Data Platform},
  author = {Kinney, Rodney and Anastasiades, Chloe and Authur, Russell and Beltagy, Iz and Bragg, Jonathan and Buraczynski, Alexandra and Cachola, Isabel and Candra, Stefan and Chandrasekhar, Yoganand and Cohan, Arman and Crawford, Miles and Downey, Doug and Dunkelberger, Jason and Etzioni, Oren and Evans, Rob and Feldman, Sergey and Gorney, Joseph and Graham, David and Hu, Fangzhou and Huff, Regan and King, Daniel and Kohlmeier, Sebastian and Kuehl, Bailey and Langan, Michael and Lin, Daniel and Liu, Haokun and Lo, Kyle and Lochner, Jaron and MacMillan, Kelsey and Murray, Tyler and Newell, Chris and Rao, Smita and Rohatgi, Shaurya and Sayre, Paul and Shen, Zejiang and Singh, Amanpreet and Soldaini, Luca and Subramanian, Shivashankar and Tanaka, Amber and Wade, Alex D. and Wagner, Linda and Wang, Lucy Lu and Wilhelm, Chris and Wu, Caroline and Yang, Jiangjiang and Zamarron, Angele and Van Zuylen, Madeleine and Weld, Daniel S.},
  publisher = {arXiv},
  year = {2023},
  doi = {10.48550/ARXIV.2301.10140},
  url = {https://arxiv.org/abs/2301.10140}
}
```

## Dataset Card Contact

Cesar Berrospi Ramis [@ceberam](https://github.com/ceberam)