LingoLift: Supporting Educators in Personalized Oral Language Teaching for Autistic Children through Content Generation

Paper Title

LingoLift: Supporting Educators in Personalized Oral Language Teaching for Autistic Children through Content Generation

Publication Info

• Topic area: Personalized oral language teaching for autistic children using AI-generated content.
• Keywords: Autism, oral language teaching, generative AI, personalized education, VB-MAPP, thematic learning, inclusive education, AI-assisted teaching, special education, AR projection.

Background and Problem

• Problem / challenge: Educators face significant challenges in creating personalized, coherent, and engaging oral language teaching materials for autistic children due to the time-intensive nature of preparation and the lack of appropriate resources. Existing technologies often overlook the role of educators or fail to address the diverse oral language needs of autistic children.
• Significance: Personalized oral language teaching is critical for improving the quality of life and reducing social isolation for autistic children. Addressing this challenge can enhance teaching efficiency and improve learning outcomes.
• Motivation and related work: Prior research has explored VR, AR, and serious games for language learning but often treats children as independent users, neglecting educators' roles. Generative AI has shown promise in creating customized educational content but lacks integration with standardized assessment frameworks and does not fully support one-on-one oral language teaching for autistic children.

Solution

• Proposed approach: LingoLift, a generative AI-powered system, supports educators in creating personalized, interest-based, and ability-adapted oral language teaching materials for autistic children. It integrates VB-MAPP for assessment-driven content generation and provides seamless lesson delivery through AR projection.
• Novelty:
  1. Integration of VB-MAPP for assessment-driven personalized content generation.
  2. Thematic coherence across multiple language skill domains (articulation, vocabulary, grammar, conversation).
  3. End-to-end workflow supporting lesson preparation, delivery, and progress tracking.
  4. Inclusion of educators as active participants, enabling creative adaptations and real-time contextual support.
• Procedure and key techniques:
  • Child Profile Management: Teachers input language abilities and interests, with automatic updates based on progress.
  • Lesson Preparation: AI generates personalized learning objectives, materials, and thematic lesson plans.
  • Lesson Delivery: AR projection enables interactive teaching, with gesture-based controls for real-time adjustments.
  • System Architecture: Combines a teacher tablet client, backend services using retrieval-augmented generation (RAG), and a classroom projection client for seamless integration.

Results

• Concrete findings:
  • Reduced lesson preparation time by 66% (from 72 minutes to 24.7 minutes on average).
  • Significant improvement in children's engagement over three weeks (p < 0.05).
  • Teachers rated learning objectives as well-matched to children’s abilities (M = 5.27, SD = 0.52 on a 7-point scale).
• Advantage over baselines:
  • Enhanced thematic coherence and personalized content compared to traditional methods.
  • Streamlined manual processes and enriched teaching materials.
  • Enabled real-time adjustments and creative pedagogical innovations.
• Experiments / evaluation:
  • Three-week deployment study with 10 teacher-student dyads (30 lessons total).
  • Mixed-methods evaluation combining questionnaires, interviews, video observations, and system-generated content analysis.
  • Participants: Children aged 6–9 with autism, assessed using VB-MAPP, and teachers with an average of 4.9 years of experience.
• Limitations and future work:
  • Deployment limited to well-resourced urban schools; further studies needed in diverse socioeconomic contexts.
  • Short-term study duration; longitudinal research required to assess long-term impact.
  • System design tailored to Chinese linguistic and cultural contexts; adaptation needed for other languages and cultures.

Summary

LingoLift is a generative AI-powered system designed to support educators in personalized oral language teaching for autistic children. By integrating VB-MAPP for assessment-driven content generation and providing thematic coherence across language domains, it reduces preparation time, enhances teaching efficiency, and improves student engagement. A three-week field deployment demonstrated its usability and effectiveness, while also revealing opportunities for creative teacher adaptations and multi-sensory compensations. Future work should explore long-term impacts, broader deployment contexts, and cross-cultural adaptations.