Autism Children Education Knowledge Graph: Construction and Validation

Children with Autism Spectrum Disorder (ASD) require specialized educational strategies grounded in interdisciplinary knowledge, yet such knowledge is often scattered across fragmented literature, limiting its accessibility and utility. In this paper, we propose a domain-specific knowledge graph (KG) for ASD education that systematically encodes factual, evidence-based information into a structured and semantically rich representation. By reviewing a broad corpus of open-source research articles and books, we construct a multi-layered ontology that captures intervention plans, practical considerations, symptom types, example cases, and communication strategies. Specifically, we extract 3,997 entities and 3,539 triples to form a high-quality KG with refined structure and minimal noise. To evaluate its impact, we develop a new dataset combining literature-derived questions with GPT-4o-generated and expert-validated answers. Extensive experiments on various LLMs show that incorporating the KG consistently improves performance, with Qwen3-0.6B achieving 82.0% enhanced responses and a maximum score gain of 2.5. Finally, we integrate the ASD educational KG into an interactive prototype educational QA system and initiate a collaboration with the National Special Education Resource Center for Children with Autism, aiming to drive future development of intelligent and adaptive educational tools for children with ASD.