Concept Generation through Knowledge Operation with LLM and Vector Database

This study proposes a concept generation method through operations based on the semantic distance between knowledge using a vector database of design knowledge and a large language model. In addition, this study develops a method to evaluate the effectiveness of the proposed method objectively through quantitative evaluation of the generated concepts in terms of novelty and usefulness. The proposed method is applied to concept generation in architectural design, using Christopher Alexander’s A Pattern Language, which is a collection of architectural knowledge. The effectiveness of the proposed method is evaluated through comparison with human-generated concepts.