This paper reports on the participation of MDMapper in the Ontology Alignment Evaluation Initiative (OAEI) 2025. MDMapper is an ontology matching system tailored for Product Master Data Model (PMDM) contexts, integrating hierarchical reasoning and property-based similarity to support both equivalence and non-equivalence correspondences. Following its initial participation in OAEI 2024 (Anatomy and Conference tracks), MDMapper also compete in these two tracks in 2025 and additionally participated in the new Beyond Equivalence track, which evaluates alignment relations beyond equivalence. Although no significant architectural changes were made to the system, MDMapper achieved outstanding performance in Beyond Equivalence track, particularly on PMDM tasks.
CEUR ISSN1613-0073
MDMapper [
MDMapper first participated in OAEI 2024 [
Track
Year Prec. Rec.
Furthermore, the newly introduced Beyond Equivalence track provided the matching context involving PMDM and STROMA/TaSeR tasks. Here, MDMapper achieved the highest scores in the PMDM tasks, such as TIM–ECLASS, ECLASS–UNSPSC, ECLASS-GPC, GPC–UNSPSC, and GPC–UNSPSC+. The results demonstrate its outperforming capability in matching classification ontologies with hierarchical structure and attribute-rich context.
Table 1 shows that MDMapper achieved stable results on the Anatomy track compared with
2024. This year, MDMapper achieved slightly lower scores on the Anatomy track due to the
inclusion of explicit correspondences beyond equivalence, which were excluded in the last year.
The precision decreased slightly from 0.926 to 0.899, and recall remained nearly unchanged,
yielding an F1-score of 0.889 in 2025 versus 0.903 in 2024. In terms of F1-measure, MDMapper
(0.889) ranked 6th of 12 compared matchers, following Matcha [
However, it is important to note that this alignment includes non-equivalence correspondences, which are evaluated against a reference alignment that considers only equivalences. This mismatch can lead to an underestimation of performance, as some correct non-equivalence correspondences are are incorrectly penalized, contributing to the observed decrease in scores compared to last year.
The Beyond Equivalence track was newly introduced in OAEI 2025 to evaluate ontology
matching systems on their ability to detect relations beyond equivalence. The datasets used in this
track are derived from Product Master Data Models (PMDMs) as well as from the STROMA [
In addition to traditional metrics, the Beyond Equivalence track employed the isAmong
evaluation [
MDMapper achieved the best results on all PMDM tasks under both traditional and isAmong evaluation methods, while its performance on the STROMA/TaSeR datasets was moderate compared to other systems. Nevertheless, as shown in Tables 1 and 2, the overall scores remain low, highlighting the challenges of the Beyond Equivalence track and the methodological gaps in current ontology matching systems.
MDMapper demonstrated robust and adaptable performance across both classical equivalenceonly and beyond-equivalence matching tasks. Its consistent results on the Anatomy and Conference tracks confirm the stability of the framework, while its performance on the Beyond Equivalence track highlights its strength in handling hierarchical and attribute-rich ontologies, particularly for PMDMs. However, the overall low scores observed across beyond-equivalence tasks underscore the persistent dificulty of identifying non-equivalence correspondences.
Future work will focus on integrating Large Language Models (LLMs) to address the challenge
of relation typing [
During the preparation of this work, the authors used Grammarly in order to grammar and spell check, and improve the text readability. After using the tool, the authors reviewed and edited the content as needed to take full responsibility for the publication’s content.
This study was funded by Innovation Fund Denmark (grant number 2050-00004B).