All projects

On-device intelligence · Inclusive interfaces

SignSpeak Junior.

A child-friendly sentence builder with speech output and locally running word suggestions.

Status
v2.0 Android prototype
Context
Communication and language practice
Technology
JavaScript · HTML / CSS · On-device NLP · Android TTS
SignSpeak Junior mobile sentence builder with word cards and AI helper suggestions
Project interface · SignSpeak Junior

Problem

Young users can benefit from an interface that lets them construct and hear short sentences without relying on a keyboard or a cloud AI service.

Solution

The application uses large vocabulary cards, sentence playback, and an on-device helper. Its compact intent model and transition model rank possible next words and help reorder short sentences.

Architecture

  1. 01Vocabulary cards
  2. 02Local intent + transition model
  3. 03Sentence builder
  4. 04Text-to-speech
Simplified data flow through SignSpeak Junior.

My contribution

  • Built the mobile sentence builder, favorites and customizable vocabulary.
  • Combined a local intent model and transition model for context-based next-word suggestions.

Engineering challenge

The helper is packaged with the interface rather than calling a remote language model. Suggestions can use the current sentence, favorites, and recent vocabulary without sending the sentence to a cloud AI endpoint.

Results & demo

Released a v2.0 prototype APK with on-device suggestions and Android text-to-speech.

What I’d improve

Evaluate vocabulary, navigation and suggestions with educators and families before expanding the vocabulary.

A communication prototype, not a clinical product. The repository preserves an APK and extracted web bundle, not the original Gradle project. Picture mode may request online images; it is not a fully offline image library.

Project references

Next projectHotel Management

Find a page