WombatWords: counting therapy dose, live, so the therapist can focus on the child
- Client:
- Western Sydney University (Dr Elise Baker)
- Collaborators:
- WSU; Neurabuild
The problem
Speech therapists treating late-talking children have to say each target word a set number of times per session, the therapeutic "dose." Counting that by hand, mid-session, is stressful and pulls attention away from the child exactly when the child needs it most.
Our approach
Neurabuild built WombatWords using OpenAI Whisper, deployed on Cerebrium. The web app listens during a session and counts target-word usage in real time, on screen. We worked embedded with Dr Baker's team to fit the tool to the actual clinical workflow rather than the other way around. Architecture led by Dawid Loubser, development led by Graham Withey.
Outcome
Therapists get a live, on-screen count of dose during Vocabulary Acquisition and Usage for Late Talker sessions. The manual tally is gone, and with it the worry about whether the therapy is being delivered correctly, so the therapist can run engaging sessions with the child and trust the count is handled.