DOI: 10.65520/erciyesfen.1932161 ISSN: 1012-2354

MAPPING THE CONVERGENCE OF ARTIFICIAL INTELLIGENCE AND DIGITAL HERITAGE IN ARCHITECTURE: A COMPUTATIONAL TOPIC MODELING ANALYSIS OF EMERGING RESEARCH TRAJECTORIES (2016-2026)

Fazıl Akdağ
Show PDF Cite
Background/Aim: The convergence of artificial intelligence (AI) and digital heritage in architectural research has grown exponentially, yet no comprehensive computational analysis has mapped its thematic structure and evolution. This study aims to identify and temporally trace dominant research trajectories at the intersection of AI and digital architectural heritage.Methods: A dual-database systematic search (Scopus and Web of Science) was combined with Latent Dirichlet Allocation (LDA) topic modeling. From 2,347 initial records, 2,100 unique articles (2016-2026) were analyzed after deduplication. Text preprocessing and vectorization were followed by LDA modeling (k = 10 topics) using scikit-learn. Because 2026 is covered only by partial-year data (records indexed through early April 2026), it was retained for descriptive completeness but excluded from all growth and trend inferences.Results: The model identified ten distinct research topics. 3D point cloud processing formed the largest cluster (34.9%), followed by HBIM and digital documentation (17.7%), semantic segmentation (16.9%), and deep learning systems (16.8%). Temporal analysis reveals a paradigm shift from conventional photogrammetric documentation toward AI-driven automated workflows; semantic segmentation and deep learning exhibited the highest growth rates between the 2020-2022 and 2023-2025 periods.Conclusion: The findings identify a critical research-practice gap: while technical AI capabilities advance rapidly, their integration into heritage conservation policy, participatory processes, and Global South contexts remains underexplored. Ethical and epistemological dimensions of automated heritage classification are virtually absent from current discourse. This study contributes a replicable methodological framework combining bibliometric retrieval with computational text mining.

More from our Archive