Confidential — Stefan Michaelcheck Only

AI-Simulated Expert Annotation for CASE-Aligned Knowledge Graph Construction in Education

2026graph constructionnovelmethod

Yong-Sang Cho, Hyun Sook Yi, Kibum Kim, Daehyup Park, Hyerim Noh, Seungsoon Kim, Minjung Kim

https://doi.org/10.1109/acdsa67686.2026.11468193OpenAlex: W7154618491
1
URLs Found
0
Internal Citations
7
Authors
usable
Abstract Quality
GPT-5.5 Abstract Analysis

Problems Identified (4)

Scalable interoperable educational knowledge representations: Digital transformation in education creates a need for scalable and interoperable knowledge representations for AI-driven personalization and assessment.

Curriculum knowledge graph construction from standards: There is a need to construct educational knowledge graphs directly from formal curriculum standards.

Scalable interoperable educational knowledge representations: Digital transformation in education creates a need for scalable and interoperable knowledge representations for AI-driven personalization and assessment.

Curriculum knowledge graph construction from standards: There is a need to construct educational knowledge graphs directly from formal curriculum standards.

Proposed Solutions (5)

CASE-aligned multi-agent KG construction: The study proposes a CASE-aligned Multi-Agent Framework for constructing educational knowledge graphs from formal curriculum standards.

Supervisor-orchestrated specialized agents: The framework uses specialized agents coordinated by a supervisor protocol to automate analysis, relationship synthesis, and validation.

Human-context-grounded relation inference: The agents use human-annotated data as contextual ground truth to infer hierarchical, sequential, and compositional relationships among learning objectives.

CASE-aligned multi-agent KG construction: The study proposes a CASE-aligned Multi-Agent Framework for constructing educational knowledge graphs from formal curriculum standards.

Supervisor-orchestrated specialized agents: The framework uses specialized agents coordinated by a supervisor protocol to automate analysis, relationship synthesis, and validation.

Results (3)

Improved KG quality over single LLM:

Scalable foundation for curriculum-aligned KG generation:

Support for downstream education applications:

Research Domain

AI in education; educational knowledge graph construction

← Back to all papers