From Statute to Screen
How AI is Democratising Access to Law in Nigeria
Yemi A.
Chief Executive Officer, TNG
Abstract
Legal language is dense, technical, and overwhelmingly written for professionals — not citizens. This paper documents TNG's Lagos Law Translation Project, in which 70+ Lagos State statutes were translated into plain English and Yoruba using a hybrid human-AI pipeline. We detail the annotation methodology, quality assurance process, accuracy benchmarking at scale, and the constitutional and policy implications of making law legible to the public. We propose a national framework for AI-assisted legal translation across Nigeria's 36 states.
1. The Access to Justice Crisis in Nigeria
Nigeria's legal system produces a significant volume of legislation at federal, state, and local government levels. Lagos State alone maintains a corpus of over 300 active statutes, covering areas from land use and tenancy to environmental law, traffic regulation, and commercial licensing. These laws bind every resident of the state — yet the vast majority are inaccessible in any practical sense to the citizens they govern.
The barriers are multiple: language (most statutes are in dense legalese that requires professional training to parse); distribution (full official compendia are not freely available online in searchable form); and linguistic exclusion (almost no Nigerian legislation is available in any indigenous language).
Article 36 of the Nigerian Constitution guarantees fair hearing, which presupposes that citizens can understand the laws they are subject to. The reality falls far short of this guarantee. When a Lagos market trader is issued a compliance notice under the Lagos State Environmental Sanitation Law 2000, they are expected to comply with provisions they have never read and would struggle to comprehend if they did.
2. Project Design: The Lagos Law Translation Pipeline
In 2024, TNG partnered with Lagos State Government to build a system capable of translating the full Lagos State Laws corpus into plain English and Yoruba at production quality and scale.
The source corpus comprised 70+ statutes totalling approximately 1.4 million words of legal text. The translation targets were: (a) plain-language English summaries and full translations suitable for a general reader with secondary education; and (b) Yoruba translations accurate enough to be used by citizens seeking to understand their rights.
Our technical pipeline proceeded in three stages. In Stage One, we used the Meta NLLB-200 model, fine-tuned on a domain-specific Yoruba legal corpus we assembled from court records, Yoruba-language newspapers covering legal affairs, and a glossary of Yoruba legal terms developed with the assistance of senior Yoruba-speaking lawyers. In Stage Two, outputs were reviewed by a panel of eight Yoruba-speaking legal professionals who corrected terminology, idiomatic phrasing, and tonal diacritics. In Stage Three, corrected translations were reintegrated into the training pipeline to improve model performance on subsequent documents.
Figure 1: Legal Technology — April 2026
3. Results and Quality Assessment
Across the 70+-statute corpus, our pipeline achieved an average BLEU score of 62.3 for plain-English translation and 58.7 for Yoruba translation — both significantly above the baseline NLLB-200 model (43.1 and 31.4 respectively) and competitive with human translation benchmarks for legal text.
Human evaluators rated 95.2% of translated passages as accurate in meaning, with 91.8% rated as fluent and natural in the target language. Error rates were highest in sections dealing with technical financial and environmental law, where specialist vocabulary in Yoruba is underdeveloped — an expected limitation that we addressed through a curated terminology appendix.
The project was completed in 14 weeks — compared to an estimated 3-4 years for equivalent human-only translation. Cost per statute was approximately 40% of the equivalent human translation cost, with quality outcomes that met the standard required for public publication.
4. A National Framework for Legal Translation
The Lagos project demonstrates that AI-assisted legal translation is technically and economically viable at state scale. We propose a national framework that would extend this approach to all 36 Nigerian states.
The framework rests on four components. First, a National Legal Language Corpus — a centrally maintained, openly licensed dataset of Nigerian legal documents annotated in English and major indigenous languages, managed by the National Institute for Legislative and Democratic Studies (NILDS). Second, a Federal Translation API that state governments can use to translate newly enacted legislation as part of the publishing workflow, rather than as a retrospective project. Third, a Quality Assurance Network of trained bilingual legal professionals in each state who review AI outputs before publication. Fourth, a Citizens' Law Portal — a searchable, mobile-optimised interface through which any Nigerian can access translated versions of laws applicable to their state.
5. Constitutional and Policy Implications
Making the law comprehensible to citizens is not merely a matter of civic convenience — it is a constitutional imperative. The right to fair hearing enshrined in Section 36 of the 1999 Constitution is substantively weakened when citizens cannot understand the laws under which they are charged.
Several African jurisdictions have moved in this direction. South Africa's Constitution requires legislation to be published in at least three official languages. Rwanda's government portal offers laws in Kinyarwanda, French, and English. Nigeria has an opportunity to go further — not merely translating legislation into indigenous languages, but establishing translation as a mandatory step in the legislative process, with AI infrastructure to make that mandate achievable at scale.
Conclusion
The Lagos Law Translation Project proves the concept: AI can translate complex legal text at production quality and scale, at a fraction of the cost and time of human-only approaches. But the technology is only valuable if it is deployed within a governance framework that ensures accuracy, accountability, and genuine citizen access. Nigeria has the legislative, technical, and civic capacity to make its laws universally comprehensible. The question is whether it has the political will. TNG stands ready to support that effort at national scale.
References
- [1]Constitution of the Federal Republic of Nigeria (1999), Section 36 — Right to Fair Hearing.
- [2]Lagos State Government (2025). Lagos State Laws: Digital Access Initiative — Project Completion Report.
- [3]NLLB Team (2022). No Language Left Behind: Scaling Human-Centered Machine Translation. Meta AI.
- [4]Papineni, K. et al. (2002). BLEU: A Method for Automatic Evaluation of Machine Translation. ACL 2002.
- [5]Sanni, A. (2021). Access to Justice and the Nigerian Legal System: Challenges and Prospects. Journal of African Law.
The Nexatlas Global
nexatlasglobal.com · Published April 2026
This paper is published under Creative Commons CC BY 4.0. You are free to share and adapt with attribution.
