LLM4OE-SLR

Large Language Models (LLMs) are increasingly being used to support ontology engineering (OE) tasks, yet a comprehensive understanding of their roles, effectiveness, and application domains remains limited. This project presents a Systematic Literature Review (SLR) focused on how LLMs contribute to various phases of ontology development. We analyzed 30 peer-reviewed papers published between 2018 and 2024 and extracted 41 distinct OE tasks where LLMs are applied.

---

Repository Structure

Paper Selection/

This folder documents the paper selection process, with each filtering step tracked in individual Excel sheets for full transparency and reproducibility.

The screening process includes the following steps:

1. Raw Search
   Initial search results retrieved from five major academic databases (Web of Science, ACM Digital Library, IEEE Xplore, Scopus, and Google Scholar), yielding 11,985 records based on predefined keywords, time range, and language filters.

2. Duplicate Removal
   Automatic deduplication across databases reduced the dataset to 5,275 unique entries.

3. Title Screening
   Manual title screening was performed, narrowing the pool to 204 potentially relevant papers.

4. Abstract & Peer-Review Filtering
   We applied two filters:

   • Peer-reviewed status
   • Abstract relevance to LLM-based ontology engineering
     This resulted in 38 papers.
     After further inspection, 8 papers were excluded due to lack of alignment with our scope (marked with strikethrough in the spreadsheet), leading to 30 final included studies.

5. Final Inclusion
   These 30 papers form the core dataset for downstream analysis.
   From them, we extracted 41 LLM-based OE activities, summarized across the ontology development pipeline.

---

Data Extraction/

This folder includes five tables designed to support analysis our research questions (RQs):

• Full_Data_Extraction_from_30_papers.xlsx
  Dataset including article metadata such as title, authors, year, peer-reviewed status, and language.

• Data_Extraction_for_RQ1.xlsx
  Used to analyze RQ1 (Ontology development activities supported by LLMs).
  Includes activity name, task definitions, and corresponding papers.

• Data_Extraction_for_RQ2.xlsx
  Used to analyze RQ2 (How LLMs support OE activities).
  Contains roles of LLMs, types of models used, input/output formats, and whether human-in-the-loop was involved.

• Data_Extraction_for_RQ3.xlsx
  Used to analyze RQ3 (Performance evaluation).
  Includes experiment existence, evaluation methods (quantitative, qualitative, hybrid), datasets used, metrics, and reported results.

• Data_Extraction_for_RQ4.xlsx
  Used to analyze RQ4 (Application domains).
  Lists domains where LLMs have been applied, such as healthcare, education, and finance.

---

Experiment Datasets/

Furthermore, in this folder we compiled a comprehensive summary of ** experiment datasets** used in the experiments evaluation reported by the included studies.
For each dataset.

• The dataset summary table is available at:
  full_experiment_datasets_summary.xlsx

For each dataset, we list:

• Acronym and full name
• Access URL
• Application domain (e.g., Medicine, Music, Anatomy)

---

Figures/

Here we list the code sources and code used for generating the figure sin the survey

You can find the full paper selection flowchart in:

• Figures/LLM4OE_SLR_search_pipeline.pdf

Additionally, we provide a visual representation of RQ1 (LLM-supported OE activities):

• Figures/Figure_for_RQ1.pdf