KenCorpus: Kenyan Language Corpus for NLP and Machine Learning

Overview

KenCorpus was a Lacuna Fund-supported project designed to build a Kenyan-language corpus that could support NLP and machine learning tasks across major Kenyan languages. The project created text and speech data resources for Swahili, Dholuo and Luhya and produced proof-of-concept systems for speech recognition and question answering.

Project logic

1. 1

   Aim

   Create open text and speech resources to enable NLP and ML for Kenyan languages.

2. 2

   Context

   Many Kenyan languages lack curated datasets needed for reliable ML models.

3. 3

   Tasks

   Collect corpora, validate annotations, release datasets and produce demo systems.

4. 4

   Success

   Open public-domain corpora and demonstration systems used by researchers and practitioners.

Focus areas

• NLP
• Speech Data
• Machine Learning
• Low-resource Languages