EduKG: An Educational Knowledge Graph

Introduction

Recently, the use of Knowledge Graphs (KGs) has met with great success for a wide range of use cases. In particular, KGs are increasingly used to improve the explainability of machine learning models. Unfortunately, there is a lack of suitable educational KGs. Therefore, we intend to contribute to the development and provision of a new educational KG, subsequently referred to as EduKG. EduKG is a knowledge graph instantiated with students' data and a broad spectrum of curricula. EduKG is built on the basis of EducOnto, from which it inherits the rich semantics and structural constraints.

Use Cases

EduKG was built with a very clear use case in mind: making (explainable) university curriculum recommendations. Considering that EducOnto underpins EduKG by providing it with a rich semantics, EduKG is particularly well suited for conducting research at the crossroads of Explainable Artificial Intelligence (XAI) and recommender systems (RSs). However, EduKG contains a wide spectrum of information and remains generic enough to be used in a wide range of downstream tasks.

Dataset Statistics

The first table just below gives a sense of EduKG dimensions:

#Fields Of Study	#Keywords	#School Subjects	#Majors	#Specialties
12	321	15	92	13

For those who are more specifically interested in using EduKG for the purpose of making curriculum recommendations, the following table should be looked at more closely:

#Users (Students)	#Items (Curricula)	#Interactions	Sparsity
3,583	286	7,021	99.31 %

Traditionally, a KG is defined by the following statistics: the number of unique entities and relations in the graph, as well as the total number of KG triplets. The following table provides these details:

#Entities	#Relations	#KG Triplets
5,452	27	36,301

Download

EduKG can be downloaded in .ttl format at the following link: edukg.ttl

</content>