metadata

language:
  - eng
  - afr
  - nbl
  - xho
  - zul
  - sot
  - nso
  - tsn
  - ssw
  - ven
  - tso
license: cc-by-4.0
task_categories:
  - sentence-similarity
  - translation
pretty_name: The Vuk'uzenzele South African Multilingual Corpus
tags:
  - multilingual
  - government
arxiv: 2303.0375
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
  - config_name: nbl-nso
    data_files:
      - split: train
        path: nbl-nso/train-*
      - split: test
        path: nbl-nso/test-*
dataset_info:
  - config_name: default
    features:
      - name: nbl
        dtype: string
      - name: nso
        dtype: string
      - name: score
        dtype: float64
      - name: __index_level_0__
        dtype: int64
    splits:
      - name: train
        num_bytes: 128131
        num_examples: 315
      - name: test
        num_bytes: 31826
        num_examples: 79
    download_size: 113394
    dataset_size: 159957
  - config_name: nbl-nso
    features:
      - name: nbl
        dtype: string
      - name: nso
        dtype: string
      - name: score
        dtype: float64
      - name: __index_level_0__
        dtype: int64
    splits:
      - name: train
        num_bytes: 128131
        num_examples: 315
      - name: test
        num_bytes: 31826
        num_examples: 79
    download_size: 113394
    dataset_size: 159957

The Vuk'uzenzele South African Multilingual Corpus

Github: https://github.com/dsfsi/vukuzenzele-nlp/

Zenodo:

Arxiv Preprint:

Give Feedback 📑: DSFSI Resource Feedback Form

About

The dataset was obtained from the South African government magazine Vuk'uzenzele, created by the Government Communication and Information System (GCIS). The original raw PDFS were obtatined from the Vuk'uzenzele website.

The datasets contain government magazine editions in 11 languages, namely:

Language	Code	Language	Code
English	(eng)	Sepedi	(sep)
Afrikaans	(afr)	Setswana	(tsn)
isiNdebele	(nbl)	Siswati	(ssw)
isiXhosa	(xho)	Tshivenda	(ven)
isiZulu	(zul)	Xitstonga	(tso)
Sesotho	(nso)

Available pairings

The alignment direction is bidrectional, i.e. xho-zul is zul-xho

afr-eng; afr-nbl; afr-nso; afr-sot; afr-ssw; afr-tsn; afr-tso; afr-ven; afr-xho; afr-zul
eng-nbl; eng-nso; eng-sot ;eng-ssw; eng-tsn; eng-tso; eng-ven; eng-xho; eng-zul
nbl-nso; nbl-sot; nbl-ssw; nbl-tsn; nbl-tso; nbl-ven; nbl-xho; nbl-zul
nso-sot; nso-ssw; nso-tsn; nso-tso; nso-ven; nso-xho; nso-zul
sot-ssw; sot-tsn; sot-tso; sot-ven; sot-xho; sot-zul
ssw-tsn; ssw-tso; ssw-ven; ssw-xho; ssw-zul
tsn-tso; tsn-ven; tsn-xho; tsn-zul
tso-ven; tso-xho; tso-zul
ven-xho; ven-zul
xho-zul

Disclaimer

This dataset contains machine-readable data extracted from PDF documents, from https://www.vukuzenzele.gov.za/, provided by the Government Communication Information System (GCIS). While efforts were made to ensure the accuracy and completeness of this data, there may be errors or discrepancies between the original publications and this dataset. No warranties, guarantees or representations are given in relation to the information contained in the dataset. The members of the Data Science for Societal Impact Research Group bear no responsibility and/or liability for any such errors or discrepancies in this dataset. The Government Communication Information System (GCIS) bears no responsibility and/or liability for any such errors or discrepancies in this dataset. It is recommended that users verify all information contained herein before making decisions based upon this information.

Datasets

The datasets consist of pairwise sentence aligned data. There are 55 distinct datasets of paired sentences. The data is obtained by comparing LASER embeddings of sentence tokens between 2 languages. If the similarity is high, the sentences are deemed semantic equivalents of one another and the observation is outputted.

Naming convention:
The naming structure of the pairwise_sentence_aligned folder is aligned-{src_lang_code}-{tgt_lang_code}.csv.
For example, aligned-afr-zul.csv is the aligned sentences between Afrikaans and isiZulu.

The data is in .csv format and the columns are src_text,tgt_text,cosine_score where:

src_text is the source sentence
tgt_text is the target sentence
cosine_score is the cosine similarity score obtained by comparing the sentence embeddings, it ranges from 0 to 1

Note: The notion of source (src) and target (tgt) are only necessary for distinction between the languages used in the aligned pair, as the sentence semantics should be bidirectional. (hallo <-> sawubona)

Citation

Vukosi Marivate, Andani Madodonga, Daniel Njini, Richard Lastrucci, Isheanesu Dzingirai, Jenalea Rajab. The Vuk'uzenzele South African Multilingual Corpus, 2023

@dataset{marivate_vukosi_2023_7598540, author = {Marivate, Vukosi and Njini, Daniel and Madodonga, Andani and Lastrucci, Richard and Dzingirai, Isheanesu Rajab, Jenalea}, title = {The Vuk'uzenzele South African Multilingual Corpus}, month = feb, year = 2023, publisher = {Zenodo}, doi = {10.5281/zenodo.7598539}, url = {https://doi.org/10.5281/zenodo.7598539} }

Licence

Licence for Data - CC 4.0 BY