Kirundi News Topic Classifier

Kirundi News Topic Classifier

Sort a Kirundi news article into one of 10 topics with a small model that runs inside your browser. A static web demo you open online or from a local copy.

Made by ReneNtabana on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space is a static web page that sorts Kirundi news articles into 10 topics. The model is a fine-tuned AfroXLMR-base in int8 ONNX form, and it runs inside your browser through transformers.js.

Because the model runs locally in the page, the text you paste is processed on your own computer.

Who it suits: It suits people who work with Kirundi text, including journalists, researchers and students of African language processing. Developers can read how a small classifier is run in a browser.

Industry | Who uses it | How they use it | Example
Journalism | Newsroom staff | Paste an article and read its predicted topic | Sorting a pile of Kirundi stories by subject
Research | Language researchers | Try the model on real articles | Seeing how it handles different news topics
Education | Students | Open the page and use the example texts | Learning how a text classifier behaves
AI development | Developers | Read index.html and normalise.js | Reusing the browser setup for another model

Skip it if your text is not in Kirundi or you need topics beyond the model's ten.

Where to use it: The page runs in these places.

• A web browser, on the Space's page at Hugging Face
• A web browser on Windows, macOS or Linux, from a local copy of the files

Use case | Who | What they get back
Classify one article | Journalists | A topic label from the model
Try the example texts | Students | Sample results to learn from
Study the browser setup | Developers | A small static site with a normalisation script

How to set it up: Online you only need a browser. A local copy needs these.

• Python and pip, for the hf download command
• A web browser

1. Install the huggingface_hub package.
2. Download the Space's files into a folder with the hf command.
3. Move into that folder.
4. Open index.html in a web browser.

pip install -U huggingface_hub
hf download ReneNtabana/kirundi-news-classifier --repo-type space --local-dir kirundi-news-classifier
cd kirundi-news-classifier

How to use it: Open the Space's page, or index.html in your local copy. Paste or choose a Kirundi news text and run the classifier. The page shows which of the ten topics the model picks.

Example: You open the page, pick one of the example texts or paste a Kirundi news article, and run it. The page then shows the topic the model assigns to that article.

For newcomers and experts: If you are new, open the Space's page and start with one of the bundled example texts. That shows you what the output looks like before you paste your own article.

If you know the field, the demo runs an int8 ONNX export of AfroXLMR-base fine-tuned for 10-way topic classification, using transformers.js, and a normalise.js file prepares the text. The trade-off is a client-side model, which avoids a server but needs a one-time download of the weights.

Open the page in a browser
Paste or pick a Kirundi article
normalise.js cleans the text
The ONNX model labels its topic

Good to know: The repository has six files in all, with index.html, normalise.js, style.css and an examples.json file at the top. The README points to a separate code repository for experiments and the report.

• License: MIT
• Cost to run: none named
• Data it reaches: the model files that load into your browser. The source does not say anything else.
• Not for: languages other than Kirundi, since it is built for Kirundi news

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: static
• License: mit

Links

• Space: https://huggingface.co/spaces/ReneNtabana/kirundi-news-classifier
• Files: https://huggingface.co/spaces/ReneNtabana/kirundi-news-classifier/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.