Welcome to Ghana NLP

Ghana Natural Language Processing

Processing (NLP) of Ghanaian Languages & it's Applications to Local Problems

Get Involved

watch video

About us

Ghana Natural Language Processing

Recently, there has been a realization that language technologies have been leaving certain languages behind - including most African languages. While Google Translate offers translation services in several African languages, no Ghanaian languages are included, and even for the included languages the quality has been generally found to be quite poor. Major challenges boil down to lack of good (indeed often any) training data, as well as lack of adoption of major language technologies for local problems.

Ghana NLP is an Open Source Initiative focused on Natural Language Processing (NLP) of Ghanaian Languages, & its Applications to Local Problems. While we will begin work on Ghanaian languages initially, likely focusing on Twi due to number of speakers, our ultimate goal is for our tools to be applicable throughout the West African subregion and beyond.

Develop better data sources to train state-of-the-art (SOTA) NLP techniques for Ghanaian languages.
Contribute to adapting SOTA techiques to work better in a lower resource setting such as this one.
Build functional systems for local applications, e.g., a “Google Translate” for Twi, Ewe, Frafra, etc.

WHAT WE HAVE DONE

Our Projects

Every project here is built and sustained by the Ghana NLP community — researchers, engineers and volunteers working on the languages they speak. Everything is open and free to build on.

Ghana Speech ID

Nsanku - LLM Translation Benchmark

Nsanku ASR Benchmark

Khaya Translator App - iOS

Khaya Translator App - Android

ABENA

Khaya Translator Web App

107

DATASETS

113

MODELS

OPEN SOURCE REPOS

600

COMMUNITY MEMBERS

Research

GhanaNLP conducts research in Natural Language Processing

We build the datasets, models and benchmarks that Ghanaian languages are missing, and we publish what we find. The work is open: our corpora, models and evaluation code are released for anyone to build on, and the research is done by the community itself.

NSANKU: EVALUATING ZERO-SHOT TRANSLATION PERFORMANCE OF LLMS FOR GHANAIAN LANGUAGES

Our most recent paper benchmarks 19 open-weight and proprietary large language models translating between English and 43 Ghanaian languages — the most comprehensive evaluation of its kind so far. The headline finding is a sobering one: no model and no language reached both high performance and high consistency, so these systems are not yet reliably usable for Ghanaian language translation at scale. The benchmark is public and community-extensible. Read the paper on arXiv

Achievements

GhanaNLP participation in events

We use platforms given to us by event organizers to present trending subjects in the industry alongside discuss some projects we've been able to accomplish.

GHANA NATIONAL AI STRATEGY LAUNCH

We attended the launch of Ghana's National Artificial Intelligence Strategy, alongside government, industry and academic stakeholders shaping how AI is built and applied in Ghana. Our interest there is the one we have carried throughout: that Ghanaian languages are part of the foundation, so the systems built on this strategy are usable by everyone in the country.