Term Extraction for Domain Modeling

Kruse-Kurbach T, Lohr D, Berges MP, Kohlhase M, Moghbeli Damaneh H, Schütz M (2024)

Publication Language: German

Publication Type: Conference contribution, Conference Contribution

Publication year: 2024

Series: Offene Bildung: Durch Technologie, Transparenz und Nachvollziehbarkeit die Zukunft gestalten

Conference Proceedings Title: Proceedings of the 22. Fachtagung Bildungstechnologien (DELFI)

Event location: Fulda DE

URI: https://dl.gi.de/items/5f135f89-1d83-4a3f-ae8e-a2e2a09a413b

DOI: 10.18420/delfi2024_33

Open Access Link: https://dl.gi.de/server/api/core/bitstreams/eb04e4ea-e2e3-438a-b34b-a3f7a0c21767/content

Abstract

Addressing learners' individual needs in large lectures is challenging, especially if we want to tailor teaching materials, learning interventions, and feedback to individual students or sub-cohorts with special needs and educational biographies. AI-based adaptive learning systems (ALS) can help. Given a fine-grained domain model, we work on a system that can maintain a learner model and use it to generate learner-adaptive course materials: targeted explanations, flashcards, or quizzes from collections of learning objects that reference the domain model. In this system, the domain model consists of a knowledge graph of theories, each introducing a set of concepts and their definitions, properties, and relations.

The conceptual and semantic relations  -- i.e., the terminological dependency relation between concepts -- in the knowledge graph, together with a model of cognitive processes, shape the learner model. In our experience, given sufficient student activity in the system, the quality and coverage of the domain model are key determining factors for the quality of the learner-adaptivity of the system. 

The domain model is essential to the system and requires a large investment. Concept definitions and statements of properties and relations can be taken from course material or textbooks. In other words, the development of domain models should be computer-assisted, not only for quality assurance but also for efficiency reasons. In this paper, we present and evaluate experiments on using natural language processing (NLP) techniques, in particular term extraction, to generate term lists for domain models. 

We focus on the following research question: \emph{To what extent can automatic keyword extraction support domain model creation?} To answer this question, we extract terms from an introductory computer science course to compare the results of different automatic term extraction tools with those of manual term extraction.

Authors with CRIS profile

Theresa Kruse-Kurbach Professur für Wissensrepräsentation und -verarbeitung Dominic Lohr Lehrstuhl für Angewandte Mathematik (Wissenschaftliches Rechnen) Marc-Pascal Berges Professorship for Computer Science Education Michael Kohlhase Professur für Wissensrepräsentation und -verarbeitung Halimeh Moghbeli Damaneh Professur für Wissensrepräsentation und -verarbeitung Marcel Schütz Professur für Wissensrepräsentation und -verarbeitung

Related research project(s)

Domain and competency modelling (VoLL-KI) Von Lernenden Lernen: Ganzheitliche daten- und wissensunterstützte Hochschulbildung und deren Gestaltung - VoLL-KI Dec. 1, 2021 - Nov. 30, 2025

How to cite

APA:

Kruse-Kurbach, T., Lohr, D., Berges, M.-P., Kohlhase, M., Moghbeli Damaneh, H., & Schütz, M. (2024). Term Extraction for Domain Modeling. In Gesellschaft für Informatik e.V. (Hrg.), Proceedings of the 22. Fachtagung Bildungstechnologien (DELFI). Fulda, DE.

MLA:

Kruse-Kurbach, Theresa, et al. "Term Extraction for Domain Modeling." Tagungsband 22. Fachtagung Bildungstechnologien (DELFI), Fulda Hrg. Gesellschaft für Informatik e.V., 2024.

BibTeX: Download