R2[RML]-ChatGPT Framework
Keywords: Semantic Web, Mapping Quality, Linked data Generation, ChatGPT
TL;DR: An approach to discover quality insights related to uplift mappings using ChatGPT.
Abstract: The purpose of this paper is to explore the potential of applying Large Language Models (LLMs) in the processes involved in linked data publication, which require a high level of domain knowledge. In particular, we are interested in the semantic and syntactic correctness of data provided by LLMs, which could be used during the development of declarative uplift mappings. The R2[RML]-ChatGPT Framework is proposed, which integrates ChatGPT to gather useful quality insights on uplift mappings required in the publication of linked data. Two system experiments were conducted, which involved inputting mappings to test the correctness of returned knowledge. The semantic correctness of key ontology terms related to 50 distinct concepts were measured. Furthermore, 150 files of relevant code were automatically generated using the framework and measured for syntactic correctness. Moreover, the framework attempted to resolve invalid syntactics, which were then reassessed.
Submission Number: 6
Loading