This paper presents an approach to classifying spatial categories of space objects using their textual properties in IFC data. As a standardized data format of building information, IFC enhances the data interoperability between the heterogeneous domain software. However, there are some problems that required information is omitted due to the technical translation error or mistake by users. The other problem is that some semantic information cannot be defined in the IFC data scheme. Manually checking and modifying this information requires an amount of time and labor. The intelligent information enrichment system can facilitate the application of BIM. In this regards, this research tried to address the interoperability problems with a machine learning-based method by automatically enhancing the semantic information of the BIM model. In this paper, we focused on the semantic enrichment of space objects by using textual data in the IFC. We implemented the NLP-based classification training model, which employs the word embedding techniques. As an early phase of research, this paper conducted training experiments with variable input properties, name and area of space, from space object in Korea school buildings. The accuracy of the classification model with a single feature (name) is measured at 85.9%, which is higher than multi feature (name and area). The suggested model can be expanded to more variable properties and target objects as a part of the semantic enrichment system.