Ìmọ̀dòtun AI
Preserving Wisdom. Empowering Generations.
Six core AI-powered features — from intelligent OCR and multi-language translation to semantic search and Traditional Knowledge labeling — working together as a complete indigenous knowledge preservation pipeline.
Each feature is designed to work independently and as part of the automated end-to-end archiving pipeline
Every uploaded manuscript flows through this pipeline automatically — only publishing once it has passed human review
The heart of the platform — a permanent, searchable digital vault for all forms of indigenous knowledge. Every archived item is tagged with language, category, community of origin, Traditional Knowledge labels, and publication date, making the entire corpus discoverable and filterable.
AI-powered translation across English and all major Nigerian languages, with a custom post-processing glossary layer that corrects AI outputs for culturally specific terms, place names, and indigenous concepts that generic translation models often mistranslate. The glossary is editable by admins and grows with every review.
Many invaluable indigenous knowledge manuscripts exist only as physical documents — handwritten, typed, or photographed. Ìmọ̀dòtun AI's OCR engine extracts text from these materials automatically, detecting whether a file needs OCR or can be parsed directly, and reporting a confidence score so reviewers know where to focus corrections.
After text is extracted, the AI analyses the content and automatically suggests the most appropriate knowledge category, the likely originating community, and a set of Traditional Knowledge (TK) labels. Admins review and approve or correct these before the item is published — maintaining ≥ 90% TK label accuracy as a platform benchmark.
A traditional keyword search would miss most of what the archive holds — indigenous concepts rarely match modern English search terms. Ìmọ̀dòtun AI's semantic search understands context and meaning, returning relevant results even when the exact words aren't present. Search for "healing plants" and find a manuscript about "àgbò herbal medicine."
Not a policy document — a built-in workflow. Before any item can be uploaded, the admin must confirm community consent. Before it can be published, a reviewer must confirm TK labels and cultural accuracy. Before it appears publicly, a Super Admin can make a final call. 100% cultural integrity is a benchmark, not an aspiration.
A transparent view of the technology stack, AI services, and performance targets behind each platform feature
| Feature | Technology / Service | Performance Target | Input Formats | Human Review |
|---|---|---|---|---|
| Digital Archive | MySQL + PHP PDO + FULLTEXT index | F1 ≥ 0.80 | All types | |
| Multi-Language Trans. | Google Translate API + custom glossary | BLEU ≥ 25 / BERT ≥ 0.85 | Extracted text | |
| Smart OCR | Tesseract / Google Vision API | ≥ 85% page confidence | PDF, JPG, PNG, TIFF | |
| AI Categorization | NLP keyword analysis + taxonomy matching | TK coverage ≥ 90% | Extracted text | |
| Semantic Search | MySQL FULLTEXT + vector similarity (Phase 5) | Relevance ≥ top-5 | All archive content | |
| Cultural Integrity | Built-in workflow gates + consent DB fields | 100% coverage | All uploads |
Everything you need to know about how the platform works, who can access it, and how content is protected
Only authenticated admin accounts can upload content. Admin accounts are created by the Super Admin (Ede Polytechnic staff). Public visitors can browse and search the archive but cannot upload.
The platform accepts PDF, DOCX, DOC (native digital documents), and JPG, PNG, TIFF (scanned images and photographs). Maximum file size is 50MB per upload.
Every item goes through a human review stage before publication. The reviewer can edit the AI-generated translation, add corrections, and flag any culturally incorrect interpretations. The corrected version is stored separately and used as the published output.
Yes — the public archive is open to anyone, worldwide, without requiring a login. Visitors can browse, filter, search, and read all published items. Only the admin dashboard and upload functions require authentication.
Consent is captured at the upload stage — the admin must confirm that the originating community has given permission for the material to be archived and published. This confirmation is stored in the database with every record and is visible on each public archive entry.
Yes. Admins and Super Admins can unpublish or delete any item at any time — for example, if a community withdraws consent or if an error is discovered. The platform supports full lifecycle management of archived content.
Try the live AI Translator, browse the public archive, or contact us to learn more about the platform.