Lanfrica – the hub for African language resources

Chris Chinenye Emezue, Ebuka Ezeike, Jo Havemann2022 · DOI

Abstract

Chris Emezue is a researcher and the founder of Lanfrica, an online resource centre that catalogues, archives and links African language resources in order to mitigate the difficulty encountered in discovering African works. He shares the vision and achievements of Lanfrica with Jo on this podcast.

Jo Havemann and Chris Emezue discussing Lanfrica.

Introduction

In the global Natural Language Processing (NLP) landscape, many African languages are frequently categorized as "low-resource." However, this label often masks a deeper underlying issue: a discovery problem. While linguistic data and research exist, they are frequently fragmented across personal websites, obscure repositories, or local university archives. This lack of centralization makes it difficult for researchers to build upon existing work, often leading to redundant efforts or project abandonment.

Lanfrica serves as a centralized, community-driven hub designed to solve this "discovery gap." By acting as a "hub of hubs," it provides a single entry point to find, archive, and link African language resources, thereby mitigating the fragmentation that hinders technological development for the continent.

Methodology and Functional Design

Lanfrica operates not as a primary storage site, but as a meta-repository focused on metadata and connectivity. Its functional methodology includes:

  • Cataloging and Linking: The platform creates a searchable index that points to the original locations of datasets, papers, and software. This ensures that resources remain with their original creators while becoming globally discoverable.
  • Community Integration: The project is deeply integrated with grassroots research collectives like Masakhane. This ensures the platform is built by and for the community it serves, reflecting the actual needs of African NLP researchers.
  • Ethical Data Governance: Rooted in the GIDA (Global Indigenous Data Alliance) CARE principles, Lanfrica emphasizes Collective Benefit, Authority to Control, Responsibility, and Ethics. This approach ensures that data usage respects the sovereignty of the source communities.

The Lanfrica Meta-Repository

As a meta-repository, Lanfrica indexes existing datasets for hundreds of African languages. It serves as a gateway to resources that would otherwise remain siloed in various NLP benchmarks or linguistic archives. By providing a unified interface, it transforms a collection of disparate data points into a coherent ecosystem for language technology development.

Key Insights: From Data Providers to Architects

A central insight from Chris Emezue’s work is that "low-resource" status is often a symptom of poor indexing rather than a lack of linguistic activity. Improving visibility can be as impactful as new data collection efforts.

Chris Emezue, founder of Lanfrica.

Furthermore, Lanfrica represents a significant shift in the AI research paradigm. Researchers from the Global South are moving from being mere "data providers" to becoming the architects of infrastructure and models. Emezue’s background—spanning reinforcement learning at Mila and ML advocacy at Hugging Face—illustrates the intersection of cutting-edge ML theory and community-focused data engineering necessary for the future of African NLP.

Results and Impact

The primary achievement of this initiative is the launch of lanfrica.com, a functional online resource center. Key results include:

  • Centralization: Successfully linking researchers to African NLP works and reducing the difficulty of discovery.
  • Ecosystem Support: Providing a hub that supports the broader goals of promoting indigenous languages in speech and language processing tasks.

Limitations and Future Work

While successful, the platform faces challenges in scaling metadata collection, currently relying on manual or semi-automated entry. Future work involves detailing quality verification processes for linked datasets. Additionally, ongoing research at the Mila Quebec AI Institute in "structure learning" and "causal inference" suggests that future iterations of Lanfrica may incorporate advanced automated discovery and data relationship mapping.

Summary updated

Abstract

Chris Emezue is a researcher and the founder of Lanfrica, an online resource centre that catalogues, archives and links African language resources in order to mitigate the difficulty encountered in discovering African works. He shares the vision and achievements of Lanfrica with Jo on this podcast.
BibTeX
@misc{emezue2022lanfrica,
  author = {Chris Chinenye Emezue and Ebuka Ezeike and Jo Havemann},
  title = {{Lanfrica – the hub for African language resources}},
  year = {2022},
  doi = {10.21428/51e64700.a187c146},
}
Figure