This poster paper describes the ongoing research project for the creation of a use-case-driven Knowledge Graph resource tailored to the needs of teaching education in Knowledge Graphs (KGs). We gather resources related to KG courses from lectures ofered by the Semantic Web community, with the help of the COST Action Distributed Knowledge Graphs and the interest group on KGs at The Alan Turing Institute. Our goal is to create a resource-focused KG with multiple interconnected semantic layers that interlink topics, courses, and materials with each lecturer. Our approach formulates a domain KG in teaching and relates it with multiple Personal KGs created for the lecturers. Poster Page (GitHub): https://naiayti.github.io/TeachingKnowledgeGraphs.io/ Is there really a problem? Teaching courses at the university level can be a demanding task; the field is radically evolving with topic trends changing each year, the material needs to be up-to-date, and finding teaching resources for a topic outside of one's main expertise can be challenging. The challenge arises both from not having a central hub, which can assist in accessing educational resources, topics, and courses, and also from the pluralistic naming and nonuniform format of similar resources across diferent institutions. Moreover, in the courses ofered by the semantic web community, we observe that equivalant courses cover diverse topics, and these courses often highly difer in naming, such as “Web AI” and “Knowledge Engineering”. Also, the educational material for lectures and labs is found in a large variety of modalities. From the COST DKG Workshop “Teaching Knowledge Graphs", Malaga, September 2023, it was identified the necessity for providing a resource that could highlight the topics and material taught in KG courses and a platform in which lectures could connect and exchange materials. Therefore, the proposal for a teaching Knowledge Graph (KG) resource was highlighted. Who cares? The primary beneficiaries are the lecturers who teach KG courses, and students who follow these courses. Those can benefit from a teaching KG in multiple use cases, as
indicated in Table 1, as the resource is designed with the intention of supporting teaching and learning in the discovery of related topics, courses, datasets and educational materials related to KGs. By enhancing the teaching process, the lecturers can leverage the resource to better explain sub-concepts, demonstrate practical applications, and provide richer educational experiences. Additionally, the students’ learning is enhanced as they can have access to a central hub with well-structured information about KG courses, and leverage similar content to their course to gain a broader perspective about their study. Subsequently, the teaching KG resource can function as a digital library with content quality assurance, where educational materials are organized by level and topic, and interconnected with the lecturers.
Moreover, the Semantic Web (SW) community needs a methodology for creating an educational teaching KG. This need is two folded; at first, a methodology generated from the community will benefit educators and practitioners. A SW framework for creating educational teaching KGs will grant educators access to the latest advancements that utilize the state-ofthe-art standards, and fully explore the capabilities ofered by the SW. On the other hand, as education is evolving, the demand for semantic solutions is prominent highlighting the need for the involvement of the SW into the theoretical foundation of learning and teaching applications.
know, we are the first working towards a
teaching KG for KG courses. However, our
contribution does not rely only on this. As
Semantic Web technologies have been widely
used in the education domain, ontologies are
commonly used for semantically enhancing
e-learning applications [
However, none of these approaches goes beyond the factual knowledge representation and limits each resource description to highlevel connections, such as which resource is taught first and who is the resource’s author. To the best of our knowledge, we are the first to aim to extract knowledge from each resource, such as topics from educational material, and represent them as new knowledge in the KG. This is significantly important, as one of our main contributions lies in defining a new way of creating educational KGs. We achieve this by building a use-case-driven and resource-focused KG that ofers a novel representation of the data it contains. The new representation goes beyond the factual knowledge and extraction of metadata, and further includes the statistical analysis and document information retrieval outcomes as part of the KG entities and properties. An example of this case is applying topic modelling to lecture notes [16], and enhancing the teaching KG with new entities as subtopics extracted from the lecture notes.
towards a big, multi-lingual, multi-modal, and high quality content educational teaching KG
that can serve as the basis of future educational applications in AI and beyond [17]. As KGs can
assist AI applications in education to ofer more personalised and better learning experience to
the learners [
Also, as quality education is underlined as a Sustainability Goal [18], raises the duty of setting standards for educational Semantic Web applications and KG resources in education. Consequently, we guarantee the quality of our resource by gathering topic descriptions and courses from the experts who teach them.
Fine, but how are you going to make it? We reuse Semantic Web vocabularies, and standards to develop the teaching KG [19]. Our resource aims to interlink topics, courses, and educational material in order to enable teachers to enhance their courses and help students access multiple resources relevant to their learning goals. Driven by the needs of the users and adding granularity to the representation of material in our teaching KG resource, we aim to contribute both to the Semantic Web advancements as well as set the basis for advanced learning analytics applications.
The teaching KG consists of two parts, the domain model and the user model, which are
developed as extensions of the EduCOR ontology [
Aren’t there any limitations? There are plenty of limitations to consider when building teaching KGs. Firstly, not all materials are open source and accessible, as frequently educational content stays hidden behind paywalls or requires institutional access. To address this, we develop the PKGs so the users can access the specific materials from each instructor and gain access to that knowledge. Moreover, the diversity in educational material and non-standardized course descriptions make automation the biggest challenge. The material and courses are present in various formats, which make their integration, interlinking, and maintenance a non-trivial task that we aim to tackle by parsing the text that can be extracted from each resource using symbolic and sub-symbolic techniques [21].
In Figure 1, we provide a visu- Topic
alisation of the multiple layers
(topic, lecture, lecturer, course,
educational material) present in the Lecture
teaching KG, in the top part of the
Figure, and a paradigm for
assisting content retrieval based on se- Lecturer
mantic similarity, in the bottom
part. For the latter, we follow
techniques similar to literature [
Moreover, we foresee a hierar- Lecture Lecture chical structure in the topics level. Week 4 hasTopic hasTopic Week 3 As topics can be classified to gen- Figure 1: A visualisation of the teaching KG components eral as hypernyms, such as “Data (top) and application (bottom). On the top is the Representation”, and more speci- topic ontology, which contains topics and subifed as subtopics, such as “RDFS topics (squares). The topic ontology is interconrepresentation”. We can introduce nected with week lectures (circles), which are part the relation of prerequisite among of a course (circles and edges in a single color). topics via statistical analysis of the Each course has a lecturer (human icon). Each available resources. The prerequi- lecture has one or many educational material (the sites are extracted by the course or- content in the circles). der of lectures and the topics those materials cover. For example, the topic of RDF based on statistical analysis of the gathered material it would be a prerequisite of the SPARQL topic.
We thank Axel Polleres for the name inspiration. This work was supported by COST Action Distributed Knowledge Graphs and the interest group on Knowledge Graphs at The Alan Turing Institute. E.I. would like to thank her grandma for the eternal love. 2023.00016. [11] H. Abu-Rasheed, C. Weber, M. Dornhöfer, M. Fathi, Pedagogically-informed implementation of reinforcement learning on knowledge graphs for context-aware learning recommendations, in: Responsive and Sustainable Educational Futures - 18th European Conference on Technology Enhanced Learning, EC-TEL 2023, Aveiro, Portugal, September 4-8, 2023, Proceedings, volume 14200 of Lecture Notes in Computer Science, Springer, 2023, pp. 518–523. URL: https://doi.org/10.1007/978-3-031-42682-7_35. doi:10.1007/978-3-031-42682-7\_35. [12] S. Kabongo, J. D’Souza, S. Auer, Orkg-leaderboards: a systematic workflow for mining leaderboards as a knowledge graph, Int. J. Digit. Libr. 25 (2024) 41–54. URL: https://doi. org/10.1007/s00799-023-00366-1. doi:10.1007/S00799-023-00366-1. [13] Q. U. Ain, M. A. Chatti, P. A. M. Kamdem, R. Alatrash, S. A. Joarder, C. Siepmann, Learner modeling and recommendation of learning resources using personal knowledge graphs, in: Proceedings of the 14th Learning Analytics and Knowledge Conference, LAK 2024, Kyoto, Japan, March 18-22, 2024, ACM, 2024, pp. 273–283. URL: https://doi.org/10.1145/ 3636555.3636881. doi:10.1145/3636555.3636881. [14] E. Ilkou, Personal knowledge graphs: Use cases in e-learning platforms, in: Companion of The Web Conference 2022, Virtual Event / Lyon, France, April 25 - 29, 2022, ACM, 2022, pp. 344–348. URL: https://doi.org/10.1145/3487553.3524196. [15] P. Chakraborty, S. Dutta, D. K. Sanyal, S. Majumdar, P. P. Das, Bringing order to chaos: Conceptualizing a personal research knowledge graph for scientists, IEEE Data Eng. Bull. 46 (2023) 43–56. URL: http://sites.computer.org/debull/A23dec/p43.pdf. [16] I. Vayansky, S. A. Kumar, A review of topic modeling methods, Information Systems 94 (2020) 101582. [17] E. Ilkou, M. Galletti, D. Dobriy, Edumultikg attains 92% accuracy in k-12 user profiling!, in: Proceedings of the ESWC, volume 2043, 2023. [18] U. Nations, Sustainable Development Goal 4, quality education, 2015. URL: https://sdgs.un.
org/goals/goal4. [19] M. Poveda-Villalón, A. Fernández-Izquierdo, M. Fernández-López, R. García-Castro, LOT: an industrial oriented ontology engineering framework, Eng. Appl. Artif. Intell. 111 (2022) 104755. URL: https://doi.org/10.1016/j.engappai.2022.104755. doi:10.1016/J.ENGAPPAI. 2022.104755. [20] E. Ilkou, T. Tolmachova, M. Fisichella, D. Taibi, Collabgraph: A graph-based collaborative search summary visualization, IEEE Trans. Learn. Technol. 16 (2023) 382–398. URL: https://doi.org/10.1109/TLT.2023.3242174. doi:10.1109/TLT.2023.3242174. [21] E. Ilkou, M. Koutraki, Symbolic vs sub-symbolic AI methods: Friends or enemies?, in: Proceedings of the CIKM 2020 Workshops co-located with 29th ACM International Conference on Information and Knowledge Management (CIKM 2020), Galway, Ireland, October 19-23, 2020, volume 2699 of CEUR Workshop Proceedings, CEUR-WS.org, 2020. [22] Z. Li, L. Cheng, C. Zhang, X. Zhu, H. Zhao, Multi-source education knowledge graph construction and fusion for college curricula, in: IEEE International Conference on Advanced Learning Technologies, ICALT 2023, Orem, UT, USA, July 10-13, 2023, IEEE, 2023, pp. 359–363. URL: https://doi.org/10.1109/ICALT58122.2023.00111. doi:10.1109/ ICALT58122.2023.00111.