An NLP pipeline for isiNdebele
Personal project
isiNdebele is a low-resource language: there is far less digital text, tooling and research for it than for languages like English or French. This project explored how standard natural language processing techniques can be tailored to a language like this.
What I did
- Collected raw isiNdebele text through web scraping.
- Cleaned and prepared the text so it could be analysed.
- Applied language modelling and semantic analysis to turn the raw text into usable insights.
Why it matters to me
It was my most significant learning during my data bootcamp, and the place where my background in linguistics first met my data work.
Tools
- Python
- BeautifulSoup
- scikit-learn