PlanTL-GOB-ES/pharmaconer

---
annotations_creators:
- expert-generated
language:
- es
tags:
- biomedical
- clinical
- spanish
multilinguality:
- monolingual
task_categories:
- token-classification
task_ids:
- named-entity-recognition
license:
- cc-by-4.0
---

# PharmaCoNER

## Dataset Description

Manually classified collection of Spanish clinical case studies.

- **Homepage:** [zenodo](https://zenodo.org/record/4270158)
- **Paper:** [PharmaCoNER: Pharmacological Substances, Compounds and proteins Named Entity Recognition track](https://aclanthology.org/D19-5701/)
- **Point of Contact:** [e-mail]

### Dataset Summary

Manually classified collection of clinical case studies derived from the Spanish Clinical Case Corpus (SPACCC), an open access electronic library that gathers Spanish medical publications from [SciELO](https://scielo.org/).

The PharmaCoNER corpus contains a total of 396,988 words and 1,000 clinical cases that have been randomly sampled into 3 subsets.
The training set contains 500 clinical cases, while the development and test sets contain 250 clinical cases each.
In terms of training examples, this translates to a total of 8129, 3787 and 3952 annotated sentences in each set.
The original dataset is distributed in [Brat](https://brat.nlplab.org/standoff.html) format.

The annotation of the entire set of entity mentions was carried out by domain experts.
It includes the following 4 entity types: NORMALIZABLES, NO_NORMALIZABLES, PROTEINAS and UNCLEAR.

This dataset was designed for the PharmaCoNER task, sponsored by [Plan-TL](https://plantl.mineco.gob.es/Paginas/index.aspx).

For further information, please visit [the official website](https://temu.bsc.es/pharmaconer/).

### Supported Tasks

Named Entity Recognition (NER)

### Languages

- Spanish (es)

### Directory Structure

* README.md
* pharmaconer.py
* dev-set_1.1.conll
* test-set_1.1.conll
* train-set_1.1.conll

## Dataset Structure

### Data Instances

Three four-column files, one for each split.

### Data Fields

Every file has four columns:
* 1st column: Word form or punctuation symbol
* 2nd column: Original BRAT file name
* 3rd column: Spans
* 4th column: IOB tag

#### Example
<pre>
La S0004-06142006000900008-1 123_125 O
paciente S0004-06142006000900008-1 126_134 O
tenía S0004-06142006000900008-1 135_140 O
antecedentes S0004-06142006000900008-1 141_153 O
de S0004-06142006000900008-1 154_156 O
hipotiroidismo S0004-06142006000900008-1 157_171 O
, S0004-06142006000900008-1 171_172 O
hipertensión S0004-06142006000900008-1 173_185 O
arterial S0004-06142006000900008-1 186_194 O
en S0004-06142006000900008-1 195_197 O
tratamiento S0004-06142006000900008-1 198_209 O
habitual S0004-06142006000900008-1 210_218 O
con S0004-06142006000900008-1 219-222 O
atenolol S0004-06142006000900008-1 223_231 B-NORMALIZABLES
y S0004-06142006000900008-1 232_233 O
enalapril S0004-06142006000900008-1 234_243 B-NORMALIZABLES
</pre>

### Data Splits

| Split | Size |
| ------------- | ------------- |
| `train` | 8,129 |
| `dev` | 3,787 |
| `test` | 3,952 |

## Dataset Creation

### Curation Rationale

For compatibility with similar datasets in other languages, we followed as close as possible existing curation guidelines.

### Source Data

#### Initial Data Collection and Normalization

Manually classified collection of clinical case report sections. The clinical cases were not restricted to a single medical discipline, covering a variety of medical disciplines, including oncology, urology, cardiology, pneumology or infectious diseases. This is key to cover a diverse set of chemicals and drugs.

#### Who are the source language producers?

Humans, there is no machine generated data.

### Annotations

#### Annotation process

The annotation process of the PharmaCoNER corpus was inspired by previous annotation schemes and corpora used for the BioCreative CHEMDNER and GPRO tracks, translating the guidelines used for these tracks into Spanish and adapting them to the characteristics and needs of clinically oriented documents by modifying the annotation criteria and rules to cover medical information needs. This adaptation was carried out in collaboration with practicing physicians and medicinal chemistry experts. The adaptation, translation and refinement of the guidelines was done on a sample set of the SPACCC corpus and linked to an iterative process of annotation consistency analysis through interannotator agreement (IAA) studies until a high annotation quality in terms of IAA was reached.

#### Who are the annotators?

Practicing physicians and medicinal chemistry experts.

### Personal and Sensitive Information

No personal or sensitive information included.

## Considerations for Using the Data

### Social Impact of Dataset

This corpus contributes to the development of medical language models in Spanish.

### Discussion of Biases

[N/A]

## Additional Information

### Dataset Curators
Text Mining Unit (TeMU) at the Barcelona Supercomputing Center ([e-mail]).

For further information, send an email to ([e-mail]).

This work was funded by the [Spanish State Secretariat for Digitalization and Artificial Intelligence (SEDIA)](https://avancedigital.mineco.gob.es/en-us/Paginas/index.aspx) within the framework of the [Plan-TL](https://plantl.mineco.gob.es/Paginas/index.aspx).

### Licensing information
This work is licensed under [CC Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/) License.

Copyright by the Spanish State Secretariat for Digitalization and Artificial Intelligence (SEDIA) (2022)

### Citation Information

```bibtex
@inproceedings{,
title = "PharmaCoNER: Pharmacological Substances, Compounds and proteins Named Entity Recognition track",
author = "Gonzalez-Agirre, Aitor and
Marimon, Montserrat and
Intxaurrondo, Ander and
Rabal, Obdulia and
Villegas, Marta and
Krallinger, Martin",
booktitle = "Proceedings of The 5th Workshop on BioNLP Open Shared Tasks",
month = nov,
year = "2019",
address = "Hong Kong, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/D19-5701",
doi = "10.18653/v1/D19-5701",
pages = "1--10",
}
```
### Contributions

[N/A]

原始信息汇总

PharmaCoNER 数据集概述

数据集描述

PharmaCoNER 数据集是一个手动分类的西班牙语临床病例研究集合，源自西班牙临床病例语料库（SPACCC），这是一个收集西班牙医学出版物的开放获取电子图书馆。

数据集摘要

• 总字数： 396,988
• 临床病例数： 1,000
• 子集划分：
  • 训练集：500个临床病例，8,129个标注句子
  • 开发集：250个临床病例，3,787个标注句子
  • 测试集：250个临床病例，3,952个标注句子
• 格式： Brat 格式
• 实体类型： NORMALIZABLES, NO_NORMALIZABLES, PROTEINAS, UNCLEAR

支持的任务

• 命名实体识别（NER）

语言

• 西班牙语（es）

目录结构

• README.md
• pharmaconer.py
• dev-set_1.1.conll
• test-set_1.1.conll
• train-set_1.1.conll

数据集结构

数据实例

每个文件包含四个列：

• 第1列：单词形式或标点符号
• 第2列：原始BRAT文件名
• 第3列：跨度
• 第4列：IOB标签

数据分割

分割	大小
train	8,129
dev	3,787
test	3,952

数据集创建

数据来源

• 初始数据收集和规范化： 手动分类的临床病例报告部分，涵盖多种医学学科。
• 语言生产者： 人类，无机器生成数据。

标注

• 标注过程： 受先前标注方案和语料库启发，与执业医师和药物化学专家合作，通过迭代过程确保标注质量。
• 标注者： 执业医师和药物化学专家。

个人信息和敏感信息

• 不包含个人信息或敏感信息。

使用数据集的考虑

数据集的社会影响

• 该语料库有助于开发西班牙语的医学语言模型。

附加信息

数据集许可

• 许可： CC Attribution 4.0 International
• 版权： 西班牙国家数字化和人工智能秘书处（SEDIA）(2022)

引用信息

bibtex @inproceedings{gonzalez-agirre-etal-2019-pharmaconer, title = "PharmaCoNER: Pharmacological Substances, Compounds and proteins Named Entity Recognition track", author = "Gonzalez-Agirre, Aitor and Marimon, Montserrat and Intxaurrondo, Ander and Rabal, Obdulia and Villegas, Marta and Krallinger, Martin", booktitle = "Proceedings of The 5th Workshop on BioNLP Open Shared Tasks", month = nov, year = "2019", address = "Hong Kong, China", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/D19-5701", doi = "10.18653/v1/D19-5701", pages = "1--10", }