README.md

---
license: cc-by-nc-4.0
dataset_info:
  features:
  - name: SOURCE
    dtype: string
  - name: TARGET
    dtype: string
  splits:
  - name: train
    num_bytes: 1724952
    num_examples: 5000
  download_size: 913357
  dataset_size: 1724952
---

# Dataset Card for CLARA-MeD-5000

## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
  - [Dataset Summary](#dataset-summary)
  - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
  - [Languages](#languages)
- [Dataset Structure](#dataset-structure)
  - [Data Instances](#data-instances)
  - [Data Fields](#data-fields)
- [Dataset Creation](#dataset-creation)
  - [Source Data](#source-data)
  - [Annotations](#annotations)
  - [Personal and Sensitive Information](#personal-and-sensitive-information)
- [Additional Information](#additional-information)
  - [Dataset Curators](#dataset-curators)
  - [Licensing Information](#licensing-information)
  - [Citation Information](#citation-information)
  - [Contributions](#contributions)

## Dataset Description

- **Homepage:** [https://clara-nlp.uned.es/home/med/](https://clara-nlp.uned.es/home/med/)
- **Repositories:** [https://github.com/lcampillos/CLARA-MeD](https://github.com/lcampillos/CLARA-MeD), [https://digital.csic.es/handle/10261/269887](https://digital.csic.es/handle/10261/269887)
- **Paper:** [http://journal.sepln.org/sepln/ojs/ojs/index.php/pln/article/view/6439](http://journal.sepln.org/sepln/ojs/ojs/index.php/pln/article/view/6439)
- **DOI:** [https://doi.org/10.20350/digitalCSIC/14644](https://doi.org/10.20350/digitalCSIC/14644)
- **Point of Contact:** [Leonardo Campillos-Llanos]([email protected])

### Dataset Summary

A parallel corpus with 5000 sentence pairs of professional and laymen variants as a benchmark for medical text simplification. This dataset was collected in the CLARA-MeD project, with the goal of simplifying medical texts in the Spanish language and reducing the language barrier to patient's informed decision making.

The corpus gathers 2 subsets:

- 3800 parallel sentences (149 862 tokens) semi-automatically aligned and revised by linguists.
- 1200 parallel sentences (144 019 tokens) manually simplified by linguists.

### Supported Tasks and Leaderboards

Medical text simplification

### Languages

Spanish

## Dataset Structure

### Data Instances

For each instance, there is a string for the source text (professional version), and a string for the target text (simplified version).

```
{'SOURCE': 'Varones mayores de 18 años, diagnosticados de alopecia androgénica grados I-IV de Sinclair, que no estén realizando ningún tratamiento capilar en la actualidad ni lo hayan realizado en los últimos 3 meses.'
 'TARGET': 'Hombres mayores de 18 años, diagnosticados de alopecia androgénica (calvicie) grados I-IV de Sinclair. Estos hombres no estarán realizando ningún tratamiento para el cabello en la actualidad ni lo habrán realizado en los últimos 3 meses.'}
```

### Data Fields

- `SOURCE`: a string containing the professional version. 
- `TARGET`: a string containing the simplified version. 

## Dataset Creation

### Source Data

#### Who are the source language producers?

1. Drug leaflets and summaries of product characteristics from [CIMA](https://cima.aemps.es)
2. Cancer-related information summaries from the [National Cancer Institute](https://www.cancer.gov/)
3. Clinical trials announcements from [EudraCT](https://www.clinicaltrialsregister.eu/)

### Annotations

#### Annotation process

- 3800 sentences: Semi-automatic alignment of technical and patient versions of medical sentences. Inter-annotator agreement measured with Cohen's Kappa (average Kappa = 0.839 +- 0.076; very high agreement).
- 1200 sentences: Manual simplification technical medical sentences, both at the syntactic and lexical level. Three independent evaluators assessed the quality of the simplification by means of a 5-point Likert scale questionnaires. Human evaluators scored the simplified sentences according to grammar/fluency (average 4.8%), semantic adequacy/meaning coherence (average 4.9%) and overall simplification (average 4.3%). 

#### Who are the annotators?

- 3800 sentences: <br>
Leonardo Campillos-Llanos<br>
Adrián Capllonch-Carrión<br>
Ana Rosa Terroba-Reinares<br>
Ana Valverde-Mateos<br>
Sofía Zakhir-Puig<br>

- 1200 sentences: <br>
Rocío Bartolomé-Rodríguez<br>
Leonardo Campillos-Llanos<br>
Ana Rosa Terroba-Reinares<br>

### Personal and Sensitive Information

No personal and sensitive information was used.

### Licensing Information

These data are aimed at research and educational purposes, and released under a Creative Commons Non-Commercial Attribution (CC-BY-NC-A) 4.0 International License.

### Citation Information

- 3800 sentences: Campillos Llanos, L., Terroba Reinares, A. R., Zakhir Puig, S., Valverde, A., & Capllonch-Carrión, A. (2022). Building a comparable corpus and a benchmark for Spanish medical text simplification. *Procesamiento del lenguaje natural*, 69, pp. 189--196. 

```
@article{2022claramedcorpus,
  title={Building a comparable corpus and a benchmark for Spanish medical text simplification},
  author={Campillos-Llanos, Leonardo and Terroba Reinares, Ana R., and Zakhir Puig, Sofía, and Valverde-Mateos, Ana and Capllonch-Carri{\'o}n},
  title={Procesamiento del Lenguaje Natural},
  volume={69},
  year={2022},
  pages={189--196},
  publisher={Sociedad Espa{\~n}ola para el Procesamiento del Lenguaje Natural}
}
```

- 1200 sentences: Campillos-Llanos, L., Bartolomé-Rodríguez, R. & Terroba-Reinares, A. R. (2024) Enhancing the understanding of clinical trials with a sentence-level simplification dataset. *Procesamiento del Lenguaje Natural*, 72, pp. 31--43.

```
@article{campillosetal2024,
  title={Enhancing the understanding of clinical trials with a sentence-level simplification dataset},
  author={Campillos-Llanos, Leonardo and Bartolom{\'e}-Rodr{\'i}guez, Roc{\'i}o and Terroba Reinares, Ana R.},
  title={Procesamiento del Lenguaje Natural},
  volume={72},
  year={2024},
  pages={31--43},
  publisher={Sociedad Espa{\~n}ola para el Procesamiento del Lenguaje Natural}
}
```

### Contributions

Thanks to [Jónathan Heras from Universidad de La Rioja](http://www.unirioja.es/cu/joheras) ([@joheras](https://github.com/joheras)) for formatting this dataset for Hugging Face.