Artifical Intelligence
Fonimu
Fominu was founded by a group of enthusiasts who believe Natural Language Processing (NLP) can revolutionize access to services in Africa. The project aims to build a Swahili NLP hub providing APIs for various tasks, pretrained models, links to datasets, and opportunities for contributions. Fominu strives to become the easiest Swahili NLP ecosystem for developers, data scientists, and even those with no coding experience.
Tech Stack Used: VPS · Vercel · Postgres · ReactJS · NextJS · TypeScript · Tanstack Query · Python
- Website : https://fonimu.vercel.app/
- Figma Design: Open Here
My primary contribution was on the frontend where I:
- Developed components that interact seamlessly with AI/NLP APIs and render results effectively
- Focused on creating an intuitive user interface to make Swahili NLP tools accessible for both technical and non-technical users
- Ensured smooth integration of API responses into the UI for tasks like text analysis and entity recognition
A significant challenge was implementing the Named Entity Recognition (NER) component. The system returned an array of entities and their categories from a given sentence. Initially, when mapping the entities to highlight them with labels and colors, only the first entity aligned correctly while subsequent entities shifted out of position. To solve this, I:
- Investigated how string manipulations were altering sentence positions during mapping
- Worked with the backend engineer (Frederick), who suggested mapping and updating the sentence in reverse order
- Implemented the reverse-order mapping approach, which preserved sentence integrity and correctly aligned all recognized entities
This solution resulted in accurate and visually clear highlighting of entities, significantly improving the usability of the NER feature.
