Skip to content
DSML.KZ
Publications

Talk · IEEE International Conference on Emerging Technologies and Factory Automation · September 1, 2026

Verification of Semantic Dictionaries with Semantic Embedding and LLMs

Sebastian Heppner, Ablay Abdimalinov, Moritz Sommer, Marko Ristin, Tobias Kleinert, Hans Wernher van de Venn

Ablay Abdimalinov
Open via DOI

Abstract

Semantic interoperability remains a key challenge for automated data exchange in industrial ecosystems. While recent advances in Natural Language Processing (NLP) and Large Language Models (LLMs) have improved semantic matching, their effectiveness depends strongly on the quality of the underlying semantic descriptions. In semantic dictionaries, inconsistencies, redundancies, and insufficient definitions can therefore impair downstream interoperability. This paper investigates the use of NLP-based methods for verifying the quality of semantic dictionaries. Using ECLASS as a representative industrial dataset, we propose a modular verification concept combining embedding-based similarity with LLM-based inspection to generate an automated quality signal for prioritizing candidate issues. The contributions include (i) the identification of quality-relevant verification objectives, (ii) their formalization as modular tests, and (iii) an experimental application to ECLASS. The results show that the approach can detect issues such as redundant entries and misaligned or insufficient definitions, while also highlighting its dependence on explicit textual descriptions. Our method therefore serves as a decision-support tool for expert quality assurance rather than a fully automated solution.

Topic ModelingSemantic Web and OntologiesLexicography and Language Studiessemantic dictionarysemantic embeddingslarge language models

Related publications