ZINDI_TEST_PRIVATE_HASH_DATASET

Multilingual Sexual and Reproductive Health Dataset (ZINDI_TEST_PRIVATE_HASH_DATASET)

Dataset Summary

ZINDI_TEST_PRIVATE_HASH_DATASET is a multilingual dataset for text-based sexual and reproductive health (SRH) content. It contains aligned text pairs across nine language-country configurations, designed to support research in natural language processing (NLP), translation, and text understanding for African languages.

The dataset is split into test for each language pair. It is suitable for sequence-to-sequence tasks such as translation, paraphrasing, or text generation in the SRH domain.

The dataset will be used for evaluation of models whose scores will be displayed on the private leaderboard.

---

Languages

The dataset covers the following languages:

Language	Code
Amharic	am
English	en
Luganda	lg
Akan	ak
Swahili	sw

Each language is paired with a specific country context:

• Aka → Ghana (aka_gha)
• Amh → Ethiopia (amh_eth)
• Eng → Ethiopia, Ghana, Kenya, Uganda (eng_eth, eng_gha, eng_ken, eng_uga)
• Lug → Uganda (lug_uga)
• Swa → Kenya (swa_ken)

---

language: - am - en - sw - lg - ak

Dataset Structure

The dataset is organized into language-first folders:

Dataset Structure

aka/
└── aka_gha/
    └── test-*
eng/
├── eng_eth/
│   └── test-*
├── eng_gha/
│   └── test-*
├── eng_ken/
│   └── test-*
└── eng_uga/
    └── test-*
lug/
└── lug_uga/
    └── test-*
swa/
└─── swa_ken/
    └── test-*

Each split contains files with two columns:

• input: original SRH text
• output: target text (translated, paraphrased, or processed)

---

Dataset Details

Config	Test
Aka_Gha	164
Amh_Eth	64
Eng_Eth	145
Eng_Gha	165
Eng_Ken	77
Eng_Uga	292
Lug_Uga	126
Swa_Ken	77

---

Use Cases

ZINDI_TEST_PRIVATE_HASH_DATASET can be used for:

• Machine translation and multilingual NLP research
• Sequence-to-sequence models in the SRH domain
• Text classification or paraphrasing
• Evaluating model performance across African languages

---

Loading the Dataset

from datasets import load_dataset

# Example: load English-Ghana split
dataset = load_dataset("AiHub4MSRH-Hash/ZINDI_TEST_PRIVATE_HASH_DATASET", "Eng_Gha")
print(dataset["test"][0])

@dataset{aihub4msrh-zindi_test_private_hash_dataset,
  title={ZINDI_TEST_PRIVATE_HASH_DATASET},
  author={HASH / AiHub4MSRH},
  year={2026},
  url={https://huggingface.co/datasets/AiHub4MSRH-Hash/ZINDI_TEST_PRIVATE_HASH_DATASET}