Comprehensive Assistive Language Model (CALM) for Neurodivergent Child Language Acquisition

The Comprehensive Assistance Language Model–CALM, is a Natural Language Processing framework designed to support consistent language acquisition in children and aid their phonetic, semantic, and syntactic understanding. The model is intentionally designed to engage the child through empathy, task simplification, age-appropriate adaptation and trait-based multimodality. Most neurodivergent children struggle with language acquisition through traditional learning methods and develop a language gap that continues to grow into their adolescence and adulthood. Timely intervention through immersive learning approaches can mitigate or prevent these challenges. CALM integrates two components-a prompt-tuned LLM and a trained transformer model. The inclusion of a user report prepared by the prompt-tuned LLM on traits, age, caretaker concerns and detected condition understanding, increases the response quality and the child’s learning engagement. Layering a transformer model for contextual refinement and grammar correction over the LLM-generated response ensures coherence and improves relevance in answering the child’s query. CALM uses emojis and graphics as part of the generated text to enhance visual learning and also has the option of speech input/output through the incorporated LLM. CALM has achieved meaningful results when measured using evaluation metrics such as accuracy and response time, as well as manually validated characteristics including empathy, bias, engagement, and relevance. CALM effectively simplifies tasks, visualises learning and aligns solutions to the child’s interests.