Concept Map Assessment Through Structure Classification

Laís P. V. Vossen, Isabela Gasparini, Elaine H. T. Oliveira, Berrit Czinczel, Ute Harms, Lukas Menzel, Sebastian Gombert, Knut Neumann, Hendrik Drachsler · 2025

Read Concept Map Assessment Through Structure Classification by Laís P. V. Vossen, Isabela Gasparini, Elaine H. T. Oliveira, Berrit Czinczel, Ute Harms, Lukas Menzel, Sebastian Gombert, Knut Neumann, Hendrik Drachsler (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-22741.txt.

Introduction

Vossen and co-authors worked in 2025 on grading concept maps by their structure. Students draw such maps to show how ideas link, and teachers use them to judge depth of learning. Map structure falls into three types: spoke, network, and chain.

The team took three hundred seventeen maps and sorted each by its main structure. They then trained sorting models on counts and summary facts drawn from each map. A decision tree reached about eighty-six percent accuracy. The authors say such a tool could sit inside grading systems and give students feedback in real time, helping teachers spot shallow or strong grasp of a topic.

Catalog record

CREATORLaís P. V. Vossen, Isabela Gasparini, Elaine H. T. Oliveira, Berrit Czinczel, Ute Harms, Lukas Menzel, Sebastian Gombert, Knut Neumann, Hendrik Drachsler
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE14 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-22741

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Concept Map Assessment Through Structure Classification Laís P. V. Vossen, Isabela Gasparini, Elaine H. T. Oliveira, Berrit Czinczel, Ute Harms, Lukas Menzel, Sebastian Gombert, Knut Neumann, Hendrik Drachsler Abstract Due to their versatility, concept maps are used in various educational settings and serve as tools that enable educators to comprehend students’ knowledge construction. An essential component for analyzing a concept map is its structure, which can be categorized into three distinct types: spoke, network, and chain. Understanding the predominant structure in a map offers insights into the student’s depth of comprehension of the subject. Therefore, this study examined 317 distinct concept map structures, classifying them into one of the three types, and used statistical and descriptive information from the maps to train multiclass classification models. As a result, we achieved an 86% accuracy in classification using a Decision Tree. This promising outcome can be employed in concept map assessment systems to provide real-time feedback to the student. [orcid=0000-0002-1725-0618, [email protected], ] *1 [orcid=0000-0002-8094-9261, [email protected], url=https://www.udesc.br/professor/isabela.gasparini, ] [orcid=0000-0003-2884-9359, [email protected] ] [orcid=0000-0002-4391-7308, [email protected] ] [orcid=0000-0001-6284-9219, [email protected] ] [orcid=0000-0002-9870-1109, [email protected] ] [orcid=0000-0001-5598-9547, [email protected], url=https://edutec.science/team/sebastian-gombert/ ] [orcid=0000-0002-4391-7308, [email protected] ] [orcid=0000-0001-8407-5314, [email protected], url=https://www.studiumdigitale.uni-frankfurt.de/91614784/Prof__Dr__Hendrik_Drachsler__Wissenschaftl__Leitung ] keywords Concept maps, Decision Tree, Structure classification † † copyrightyear: 2024 † † copyright: Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). † † address: Santa Catarina State University, Joinville, Brazil † † address: Federal University of Amazonas, Manaus, Brazil † † address: IPN— Leibniz Institute for Science and Mathematics Education, Kiel, Germany † † address: Studiumdigitale & Faculty of Computer Science, Goethe University, Frankfurt am Main, Germany † † address: DIPF— Leibniz Institute for Research and Information in Education, Frankfurt am Main, Germany † † corresp: Corresponding author. † † venue: EvalLAC 2024 1 Introduction Concept Maps (CM) are a robust graphical tool for organizing and representing knowledge, and they can be used for conducting assessments Ruiz-Primo 2004 , Strautmane 2012 , or even as a study aid that facilitates the observation of knowledge and potential concepts that the student may be struggling to assimilate Kirsten E. De Ries et al. 2022 , either way, it is considered a robust test of knowledge and understanding of a topic Hay and Kinchin 2006 . The assessment methods for CM are diverse, considering both qualitative and quantitative aspects of their construction Keppens and Hay 2008 . In either form, the analysis of these maps can provide valuable insights into the student’s level of learning, highlighting strengths and knowledge gaps. One of the assessment techniques for concept maps, proposed by Kinchin et al. Kinchin et al. 2000 , involves categorizing them based on their structure and dividing them into three distinct categories: network, chain, or spoke. These structures proved to be a robust way Kinchin and Hay 2005 ; Kinchin et al. 2005 for the educator to visualize the different levels of understanding of a given topic by analyzing the concept map structure that the student created Gerstner and Bogner 2009 ; Hay and Kinchin 2006 . Hence, the relevance of concept maps in education is noteworthy, as they serve as a valuable learning instrument that aids educators in the teaching process. Therefore, categorizing the CM by their structure format is crucial to understanding students’ rationale, identifying improvement opportunities, and optimizing the learning process. However, the manual categorization process can be time-consuming or even impractical in scenarios with large volumes of produced maps. Therefore, this article addresses the following research question: Can concept maps be classified as Spoke, Chain, or Network based on their features and without relying on human visual interpretation? To answer this question, we manually classified over 300 concept maps about two distinct topics in one of the three structures. Through feature engineering, we extracted features that characterize each map and used this information to train various machine learning models to find one capable of identifying the maps into either a network, spoke, or chain structure with satisfactory accuracy. 2 Concept Maps and their Structural types The need to establish standards for evaluating these concept maps has arisen. Different evaluation approaches emerged, including quantitative, holistic, similarity-based on a model concept map, or qualitative ones Kirsten E. De Ries et al. 2022 . Among these evaluation methods, Kinchin et al. Kinchin et al. 2000 proposed categorizing concept map structures into three types: spoke, network, and chain, as illustrated in Figure 1 . Figure 1: Examples of concept maps in three structures: A) spoke; B) chain; C) network. Source: Kinchin et al. (2005) Kinchin and Hay 2005 An example of a spoke-type map is depicted in Figure 1 in panel A. This type of map exhibits a circular structure in which all the related aspects of the topic are directly linked to the core concept but not directly connected. Such a structure often indicates superficial and underdeveloped knowledge of a subject. Although new knowledge can be more easily integrated, it may only be accessible through reference to the core concept Hay and Kinchin 2006 ; Kinchin et al. 2000 . In the same figure, panel B illustrates a chain-type map. These structures lack linkage between nodes in different hierarchies and tend to move through available information from start to finish quickly. Consequently, understanding concepts in the middle of the sequence without starting from the beginning may be challenging for students. Additionally, incorporating new knowledge, especially if it disrupts an established sequence, can be difficult Hay and Kinchin 2006 ; Kinchin et al. 2000 . Finally, panel C exemplifies a network-type map, distinguished by its robust and adaptable cognitive typology. This structure enables the comprehension of a diverse range of concepts through different but related routes of thought, demonstrating students’ flexibility in understanding the concept beyond a connection to a core concept or following a specific line of thought Hay and Kinchin 2006 ; Kinchin et al. 2000 . 3 The Database A collaboration with the Leibniz Institutes DIPF and the IPN provided data from the use of a custom-made digital concept mapping tool that was used within a ten-week teaching unit Czinczel et al. ress , which was implemented with upper secondary level biology classes in the federal state of Schleswig-Holstein, Germany. Preparation for analysis involved a dataset comprising 317 distinct maps with at least two links between concepts and, therefore, at least three concepts. The maps are divided into two categories: concept maps about the system “school” and concept maps about evolution. 3.1 Data preparation and cleaning The dataset underwent a transformation process to turn the nodes and edges connections into a directed graph structure, allowing the extraction of statistical data about the number of connections which node had. There was also a feature engineering procedure whereby each map was compiled within a single row of the dataset containing the following information: • Map identification: the file name that contained the entire concept map; • Number of nodes: the sum of all nodes in the concept map; • Number of edges: the sum of all edges in the concept map; • Ratio between nodes and edges: calculated by dividing the number of edges by the number of nodes in that map. This variable indicates the overall connectivity of a concept map;

Read the complete text