A master’s research system that analyses public customer feedback about Cameroon’s mobile operators, classifies sentiment and problem type, scores severity and turns it into recommendations on a live dashboard.
The challenge
Telecom customers in Cameroon complain in English, French, Pidgin and Camfranglais, across app stores and social media. Generic sentiment tools miss most of it, so managers see volume, not priorities.
What we built
The pipeline detects the language, removes personal data, normalises Camfranglais and Pidgin with a custom lexicon, then classifies sentiment and complaint category. A severity score combines intensity, category, lexicon signals and complaint velocity, and eleven rules turn the result into actions, from outage escalation to churn-risk retention.
We joined when the thesis was at risk of not being finished, helped Ludiy Ntirpang complete a working system, and showed him how to use AI to build websites and tools. He has since founded Ludiy Tech Academy and become an AI tutor and author himself, teaching AI video generation and AI for business.
Key features
Multilingual
English, French, Pidgin and Camfranglais.
Severity scoring
What to fix first, not just what people said.
Regional heatmap
Douala, Yaoundé, Buea, Bamenda, Bafoussam and Garoua.
Privacy first
Personal data is anonymised before analysis.
What the client says
When my master’s project was at risk of not being finished, MOGADONKO stepped in, helped me build a working system and showed me how to use AI to build websites and tools. It changed my path: today I teach AI myself and run Ludiy Tech Academy.
Ludiy NtirpangFounder, Ludiy Tech Academy · Master’s research, CHITECHMA, Buea Cameroon Telecom AI Decision Support System
By the numbers
- 20,454
- Feedback records in the training set
- 0.81
- Sentiment F1 score on combined data
- 149
- Pidgin and Camfranglais lexicon terms
- 3
- Languages: English, French, Pidgin