Showcase video
Infographic
Download
About this showcase
This project modernizes phonological literacy assessments by transitioning Halifax Learning's traditionally manual, one-on-one testing into an automated, asynchronous digital platform. Designed to eliminate the operational bottlenecks of face-to-face evaluations, the system allows students to complete phoneme synthesis, analysis, and recognition tasks independently on any device.
The user-friendly frontend, built with ReactJS, ensures an engaging, child-friendly recording experience. Behind the scenes, a robust Django backend processes audio recordings asynchronously, employing a pre-trained speech-to-text transformer model to transcribe spoken responses into phonetic representations. These transcriptions are automatically compared against correct grading keys to generate immediate, objective scoring.
To ensure absolute evaluation accuracy, the system maintains a hybrid human-in-the-loop design with an intuitive educator verification dashboard. This allows instructors to rapidly review auto-graded results, saving hours of active administrative time while simultaneously building a high-quality database of human-verified evaluations to continuously fine-tune the machine learning model. Containerized using Docker, the platform offers a highly scalable and adaptable solution that seamlessly integrates into modern educational frameworks.
Additional material
At a glance
- Company Halifax Learning
- SectorEducation
- Presented byMelinda Cameron
- Research projectMitacs IT45193 — Automation of Phonological Skills Assessment with Machine Learning
- SupervisorsPawan Lingras
- UniversitiesSaint Mary's University