AI in Education · Personalized Learning
Interactive AI Storybook for Early Learners
A multimodal research platform combining personalized story reading, speech interaction, AI-supported response evaluation, and structured experimental conditions for early learners.
- JavaScript
- OpenAI
- Firebase
- Speech AI
Overview
This platform was built as part of a collaborative PhD research project examining how young children interact with different forms of digital story reading. Within one research environment, it supports learner setup, experimental-condition assignment, personalized or fixed story experiences, multimodal interaction, and post-story activities.
Technical architecture
Frontend
- JavaScript modules for story reading and personalization
- Assessment, audio, agent, and screen-flow services
AI / Speech
- OpenAI transcription and response evaluation
- Text-to-speech: OpenAI, ElevenLabs, Google, and Gemini
- Browser speech fallback
Backend / Data
- Vercel serverless APIs; Flask for local development
- Firestore and Storage for session, assessment, conversation, and audio records
- API keys remain server-side
Research and learner flow
Learner Profile
Experimental Condition
Story Interaction
Personalization where applicable
Interactive Reading
Vocabulary / Voice Interaction
Post-Story Conversation or Reflection
Session Storage
The system supports personalization, no-personalization, and expository research conditions. The personalization condition includes interactive customization and AI-supported interactions; the other conditions use fixed story experiences as specified by the research design.
Key features
Personalized research conditions
Personalized, non-personalized, and expository experiences run within a consistent workflow, with story elements customized only where the research condition requires it.
Multimodal learner interaction
Story reading can combine narration, vocabulary support, speech input, recorded responses, and interactive learner-facing elements.
AI-supported responses and conversation
Configured voice responses can be transcribed and evaluated through backend AI services, with local keyword fallback. Structured post-story conversations can provide AI-generated or predefined feedback according to configuration.
My role and technical contributions
Within the broader research project, my role focuses on the learner-facing platform: interactive story flows, personalization, AI and speech-service integration, backend response evaluation, and research-data persistence.
Interface views
Learner profile and condition setup
Researchers enter learner and session information, including the assigned study condition, before the reading experience begins.
Story personalization
Story-specific choices allow selected characters, objects, or settings to be customized for the personalized reading condition.
Interactive story reader
Illustrated story pages support narration and learner-facing interactions during the reading experience.