Akademik Veri Yönetim Sistemi
URBAN MORPHOLOGY IN THE AGE OF ARTIFICIAL INTELLIGENCE, Turin, İtalya, 16 - 20 Haziran 2025, (Yayınlanmadı)
In the urban planning process in Turkey, plans are integrated with plan notes and reports, and the plan notes produced during the plan-making process undertake a critical function in shaping the urban forms in terms of ensuring the applicability of plan decisions and their compliance with spatial features. Plan notes provide professionals with either a restrictive or detailed framework in the design process of urban forms. However, the writing of these notes is a time-consuming process that requires intensive specialized knowledge and is open to interpretation. Differences in the content and style of plan notes in different cities also make standardization difficult. This study aims to develop a GPT-based language model prototype that can support planners' plan note production processes by associating plan notes from various cities in Turkey with urban morphology indicators and fine-tuning them on OpenAI GPT models. First, plan notes were collected from different cities and these notes were matched with the urban form features (morphological indicators such as urban density, number of floors, building block typology) of the generated plan. The resulting prototype generates draft plan notes that can be used in plan-making processes based on spatial inputs received from the user (low density residential area, maximum 3 floors). Preliminary results show that the GPT model can provide planners with fast and consistent recommendations for producing plan notes supported by urban form data. In particular, providing morphological factors such as land use characteristics, city block typology and density as input data to the model has enabled the model to produce content that is more compatible with local planning decisions. This study provides an important application example in the field of urban planning and design on the use of large language models in specialized fields such as urban planning by fine-tuning them with local data.