Table 3.

A summary of selected KG construction approaches for concept mapping and visualization.

Ref.	KG Specific Purpose	Construction Algorithm(s)	Type of KB	KG Resource(s)	#Entites (e)/#Relations (r)	Evaluation Criteria	Limitation(s)
[86]	Improve learning outcomes	Ontology based concept mapping	Schema-based	Textbooks, course slides, and course syllabi	#e: 60,000 #r: 80,000	Correlation analysis, ranking entities	•Limited discussion on the KG construction algorithms. •Poor evaluation methodology.
[4]	Visual representation of the curriculum contents.	N/A	Schema-based	Chinese University MOOC, PTA, and Rain Classroom.	N/A	Average scoring rates	•Focus on K12 education disciplines, with less research in higher education disciplines. •Lack of comprehensive research on the combination of theory and practice in higher education. •Limited discussion on the KG construction and evaluation.
[90]	Visual representation of complex concepts in cybersecurity	NER and ontology mapping	Hybrid	Lecture notes, project lab manuals, quizzes, etc.	#e: 62 #r: 44	Surveys and interviews	•Potential difficulty of accurately extracting entities and relations from highly varied and complex unstructured texts. •The approach was not validated on downstream tasks.
[58]	Identifying subject teaching resources	BERT-BiLSTM-CRF	Hybrid	Teaching resources (syllabuses, textbooks, lesson plans) and internet encyclopedia texts (similar to DBpedia for expansion).	#e: 1225 #r: 1722	PMI, NGD, baselines comparison	•Inadequacy in capturing all knowledge points, •The choice of parameters such as context window size (k) and threshold values affects the graph's structure, •Potential limitations in the coverage of internet encyclopedia texts.
[91]	Knowledge building community	Mutual information, adjacent information entropy, topic modeling, association rule mining, and pattern matching	Schema-free	Students' notes, reflections, and summarizations in the field of physics subjects.	The paper does not explicitly mention the exact numbers, but it states that a total of 7339 junior middle school physics texts were collected for the dataset.	precision, recall, F1-score, and qualitative analysis	•Sensitivity to parameter settings, •Generalizability to other domains, •Scalability to larger datasets, •The challenge of ensuring high-quality educational relations through automated techniques.
[101]	Visualizing and querying medical knowledge.	Determining entities, attributes, and relationships based on user demand analysis and data sources.	Schema-based	Internet medical encyclopedia, medical encyclopedia, and other medical knowledge sources.	N/A	Clarity, accessibility, and comprehensiveness of the visualized knowledge map,	•No proper discussion on the mechanisms followed to construct the KG, •Poor evaluation techniques.
[102]	The interactive dictionary that offers descriptions, meanings, and semantic networks of programming skills	Rule-based concept mapping	schema-based	Wikipedia	N/A	A comparison of PS-Dict with other computer dictionaries.	•Lack of discussion of the accuracy of the heuristic rules used for retrieving Wikipedia articles, •No proper discussion of the mechanisms followed to construct the KG.
[88]	Liberal arts subjects	"four-step method" involving domain experts, semantic annotation, and data enrichment.	Schema-based	historical events, literary works, geographical information, and related educational content,	N/A	N/A	•The article lacks specific data or statistics regarding the size and complexity of the KGs created. •It does not explore the technical details of constructing the KG, making it challenging to replicate the process. •The integration of the KG relies on manual annotation and may benefit from more automated methods. •Practical applications and benefits for educators and students are not elaborated upon in detail.
[89]	Educational KG Construction and Management	NLP and EduLink	Hybrid	LRMI Standard for educational resources, and external data sources, such as schema.org, YAGO, Wikidata, and diverse online data.	#e:>2.5 M	A comparison with other existing educational KGs using data sufficiency metric	•Lack of specific details on construction algorithms. •The challenge of indexing and linking heterogeneous online data effectively. •The complexity of representing rhetorical roles in educational knowledge.
[103]	Internal policy control conceptualization and visualization in higher education	Adhoc		CNKI database		mean Silhouette	•Limited data sources, •limited application scope, •poor evaluation metrics, •KG embedding was not properly demonstrated and evaluated