FonoDetektor

FonoDetektor helps Slovak therapists and parents spot kids’ speech patterns by comparing target and spoken words, detecting common phonological processes with easy-to-understand results.

Comment

🧠 Inspiration

Many children in Slovakia experience phonological delays, but there are few accessible tools for analyzing their speech. We were inspired to create a friendly, Slovak-language web app that helps parents and speech therapists detect phonological processes early — using technology that's simple and available in the browser.

🗣️ What it does

FonoDetektor lets users select a Slovak word (with an image), record a child’s spoken version, and automatically detect common phonological processes such as Posúvanie artikulácie dopredu, Deafrikácia, or Kĺzanie likvíd. The app then explains each detected process with visual feedback and Slovak-language descriptions based on AFOP standards.

🛠️ How we built it

We used:

  • JavaScript + HTML for a clean, accessible web interface
  • MediaRecorder API for in-browser audio capture
  • A Whisper-based transcription API to transcribe Slovak speech
  • A set of phoneme comparison rules based on AFOP's classification of phonological processes
  • Static images mapped to test words for visual context and engagement

🧩 Challenges we ran into

  • Automatic Slovak transcription is not well supported in browser-native APIs, so we had to integrate a remote Whisper-based solution. While it works, it's not always perfectly accurate.
  • Creating a reliable phoneme mapping system and comparison logic for Slovak required custom rule-building and some manual dictionary work.
  • Sourcing enough child-friendly images for each test word and scaling them properly was also time-consuming.

🏆 Accomplishments that we're proud of

  • Developed the first interactive, Slovak-language tool for phonological process detection
  • Integrated AFOP standards directly into a user-friendly interface
  • Made the app work entirely in the browser, with no login or setup
  • Designed for both experts (SLPs) and non-experts (parents, teachers)

📚 What we learned

  • How to model phonological rules in a structured, language-specific way
  • The limitations of current browser APIs for speech-to-text, especially for Slovak
  • How crucial visuals and simplicity are when designing tools for educational or therapeutic use

🚀 What's next for FonoDetektor

  • Improve transcription quality, possibly by fine-tuning Whisper or allowing manual phoneme correction
  • Expand the image library with more target words, adjusted dynamically based on the child’s past performance
  • Add progress tracking, exportable session reports, and therapist notes
  • Let users create custom word lists for personalized therapy paths
  • Explore adding real-time phoneme highlighting as the child speaks

Built With

Share this project:

Updates