ESOL Pronunciation LMS
A dialect-aware speech coach that hears learners phoneme by phoneme, then tells them what to fix.
An ESOL (English for Speakers of Other Languages) learning platform built around automated speech analysis. Learners record themselves, and a containerized pipeline of speech-recognition, phoneme, grammar and feedback services scores their pronunciation against 17 British dialects. Lessons, progress tracking, community and educator tools sit around that core.
The Motivation
Most language apps focus on vocabulary and grammar and treat pronunciation as an afterthought. That's the skill non-native speakers say holds them back most at work and in daily life. I wanted a tool that does what a good ESOL teacher does: listen closely, break speech into its smallest sounds, compare them with a real target accent, and coach the learner on the exact sound that's wrong. I also wanted to show that a solo engineer could design it properly. That meant polyglot microservices, honest cost trade-offs and a learner experience that doesn't feel like a lab tool.
The Problem It Solves
ESOL learners get almost no objective feedback on pronunciation between classes, and 'English' isn't one target: a learner moving to Liverpool or Glasgow is aiming at a different sound system from someone learning Received Pronunciation. Speech analysis also takes a lot of compute. ASR, phoneme alignment and NLP can't run inside a web app without taking it down, and paid cloud AI adds per-second costs and sends learner audio to a third party.