تبديل اصطلاحنامه منطق پايگاه مديريت اطلاعات علوم اسلامي به هستي‌شناسي با بهره‌جويي از هوش مصنوعي

Thesaurus , Ontology , Logic , Islamic Information an‎d Documents Management Research Institute , AI Platforms , Large Language Models (LLM)

Converting the Logic Thesaurus of Information Management Database of Islamic Science into Ontology using Artificial Intelligence (AI)

علم اطلاعات و دانش شناسي

The objective of the present research was the conversion of the thesaurus of logic of the Islamic Sciences Information Management Database into an ontology by utilizing artificial intelligence. In terms of objective, the research was applied, an‎d in terms of approach, it was qualitative. The research population consisted of 6569 preferred an‎d non-preferred terms existing in the Islamic Sciences Information Management Database under the domain of logic, as well as various types of artificial intelligence platforms (with different large language models). The research sample was the vocabulary block of "Alfaz" Terms from the logic thesaurus, consisting of 235 terms, an‎d four artificial intelligence platforms: DeepSeek, Grog, ChatGPT-5, an‎d Gemini, which were selec‎ted through purposive sampling method. The data collection tool was an audit checklist based on ontology elements. The research process was designed an‎d implemented in twelve consecutive steps an‎d was based on a cycle of "optimizing pro‎mp‎t design, execution, an‎d recording results" for five main tasks (separation of concept from instance, creation of class structures, allocation of instances, extraction of relations, an‎d determination of relation properties). To create the ontology, the tool Protégé edition 5.6.5 was used. Also, to use GPT-5 artificial intelligence, coding in Python language within the Visual Studio software environment was employed, an‎d for the other three platforms (DeepSeek, Grog, an‎d Gemini), web-based user interfaces were used. The method of analysis in this research was qualitative an‎d of a structural-interpretive type. The findings showed that no platform achieved absolute an‎d integrated performance, an‎d the superiority of one model in one step (such as separation of concepts from instances) did not guarantee its success in the next step (such as creating hierarchies). Ultimately, after a dual comparative analysis of the outputs by artificial intelligence an‎d a domain expert, the outputs of the "Grog" an‎d "DeepSeek" platforms had better performance, an‎d for the final implementation in the Protégé software, the Grog output was selec‎ted based on the expertʹs opinion. It is worth mentioning that all outputs, especially in the section of constructing class an‎d subclass, required modifications. The results emphasized the necessity of final supervision an‎d validation by an expert as an indispensable factor. Ultimately, this research showed that artificial intelligence platforms can act as assistants in the complex process of ontology development, but their outputs alone lack the necessary validity an‎d accuracy for use in an ontology system. The optimal approach is adopting a hybrid an‎d collaborative approach in which artificial intelligence reduces the initial workload, an‎d the final product is shaped under the meticulous review, completion, an‎d correction of domain experts.

پژوهشكده مديريت اطلاعات و مدارك اسلامي

5

علي ميرعرب

154043