Abstract: Specification translation is an important step of BIM compliance checking, and it is also the technical basis and prerequisite for realizing automatic and intelligent code compliance checking. The first step of specification translation is to automatically classify design specifications into predefined categories for preparing the subsequent text analysis and rule extraction. However, due to the lack of corpus in the field of structural design, the automatic classification technology of design specifications needs to be developed. Therefore, based on the "Code for Concrete Structure Design" and "Code for Seismic Design of Buildings", a structural design corpus is created. According to IFC entity name catalog, an automatic classification method of structural design code is proposed by Python language programming and text classification algorithm based on machine learning. The process can be divided into three steps: data preparation and text preprocessing; feature extraction and selection; training, testing and evaluation of classifiers. The results show that the classification method can effectively realize the automatic classification of structural design specifications, and the accuracy and recall rate of the classifier to the test specifications can reach 75%and 83%.