README.md

---
language:
- ca
- de
multilinguality:
- multilingual
pretty_name: CA-DE Parallel Corpus
size_categories:
- 1M<n<10M
task_categories:
- translation
task_ids: []
license: cc-by-nc-sa-4.0
---

# Dataset Card for CA-DE Parallel Corpus

## Dataset Description

- **Point of Contact:** [email protected]

### Dataset Summary

The CA-DE Parallel Corpus is a Catalan-German dataset of parallel sentences created to support Catalan in NLP tasks, specifically 
Machine Translation.

### Supported Tasks and Leaderboards

The dataset can be used to train Bilingual Machine Translation models between German and Catalan in any direction, 
as well as Multilingual Machine Translation models.

### Languages

The sentences included in the dataset are in Catalan (CA) and German (DE).

## Dataset Structure

### Data Instances

A single tsv file is provided with the sentences sorted in the same order and 
a header containing the two-letter ISO language code for the language in each column: 
ca-de_all_2023_09_11.tsv.


### Data Fields

[N/A]
    
### Data Splits

The dataset contains a single split: `train`.

## Dataset Creation

### Curation Rationale

This dataset is aimed at promoting the development of Machine Translation between Catalan and other languages, specifically German.

### Source Data

#### Initial Data Collection and Normalization

The first portion of the corpus is a combination of the following original datasets collected from [Opus](https://opus.nlpl.eu/): 
Multi CCAligned, WikiMatrix, GNOME, KDE4, OpenSubtitles, GlobalVoices, Tatoeba.

Additionally, the corpus contains synthetic parallel data generated from the original Spanish-Catalan Europarl and Tilde corpora 
made public by [SoftCatalà](https://github.com/Softcatala/Europarl-catalan).

A last portion of the dataset is composed by synthetic parallel data generated from a random sampling of the Spanish-German corpora 
available on [Opus](https://opus.nlpl.eu/) and translated into Catalan using the [PlanTL es-ca](https://huggingface.co/PlanTL-GOB-ES/mt-plantl-es-ca) model.

All datasets are deduplicated and filtered to remove any sentence pairs with a cosine similarity of less than 0.75.
This is done using sentence embeddings calculated using [LaBSE](https://huggingface.co/sentence-transformers/LaBSE).
The filtered datasets are then concatenated to form the final corpus.

#### Who are the source language producers?

[Opus](https://opus.nlpl.eu/) 

[SoftCatalà](https://github.com/Softcatala/Europarl-catalan)

### Annotations

#### Annotation process

The dataset does not contain any annotations.

#### Who are the annotators?

[N/A]

### Personal and Sensitive Information

Given that this dataset is partly derived from pre-existing datasets that may contain crawled data, and that no specific anonymisation process has been applied, 
personal and sensitive information may be present in the data. This needs to be considered when using the data for training models.

## Considerations for Using the Data

### Social Impact of Dataset

By providing this resource, we intend to promote the use of Catalan across NLP tasks, thereby improving the accessibility and visibility of the Catalan language.

### Discussion of Biases

No specific bias mitigation strategies were applied to this dataset. 
Inherent biases may exist within the data.

### Other Known Limitations

The dataset contains data of a general domain. Application of this dataset in more specific domains such as biomedical, legal etc. would be of limited use.

## Additional Information

### Dataset Curators

Language Technologies Unit at the Barcelona Supercomputing Center ([email protected]).

This work has been promoted and financed by the Generalitat de Catalunya through the [Aina project](https://projecteaina.cat/).


### Licensing Information

This work is licensed under a [Attribution-NonCommercial-ShareAlike 4.0 International](https://creativecommons.org/licenses/by-nc-sa/4.0/).

### Citation Information

[N/A]

### Contributions

[N/A]