TRANSLATION BETWEEN LINGUISTIC STRUCTURES AND SHAPE STRUCTURES FOR BIDIRECTIONAL DESIGN
In upstream design, designers generate new images. This process is important in forming the whole design object. In this process they use verbal image and nonverbal image in their mind. We hope to support design by matching between language and shapes. For this purpose, we believe the bottom-up approach is better because designer's mind is dynamic. The Statistical Machine Translation is known in Natural Language Processing as such method. We proposed Linguistic Relational Model (LRM) to analyze verbal data, and chose CSG as nonverbal data. Then we match between LRM and CSG by statistical approach. In order to evaluate this method, we create transcription data for LRM and calculate the accuracy of our method.