Developing and Utilizing Topic Classifier Using Statistical Latent Semantic Analysis in Architectural Design

In order to utilize machine learning and generative AI to assist architectural design, it is important to visualize the relationship between the words used by designers and topics. For this reason, we developed a topic classifier using architecture magazines and conducted two investigations. First, we classified students’ architectural design texts and evaluated the classification performance. Topic classifiers have difficulty classifying sentences that contain multiple topics. Second, we presented similar cases of students design by using document vectors. Among similar cases, students focused on ideas that were new to them and solutions to their own problems.