In BIM projects, BIM Collaboration Format (BCF) files are a central resource for interdisciplinary coordination and communication between different project participants. During a project, complex information is collected in the BCF files, but this information is often insufficiently analysed. An optimized, cross-project evaluation of the different BCF files can be achieved based on fundamental classifications of the BCF files and the information contained within them. However, the existing classification options in the BCF format, especially for complex construction projects, as well as the actual classification process in the construction process, often prove to be insufficient. This is not only due to the additional effort required for classification but also to the difficulty of accessing specific information, such as building components described in the BCF. The direct linking of further information resources is difficult in the BCF format, which makes an automated classification of the BCF files even more difficult. This paper presents an innovative classification scheme that takes into account the structure of BCF files and extends it with various classification parameters. The primary goal is to extract the information from BCF files and convert it into a structured data format. This enables efficient storage in a database and provides the basis for later analysis. The classification of BCF files will be partially automated by using information extraction methods that utilize natural language processing (NLP) to extract essential text segments from the unstructured information in the BCF files. This methodology will be used to simplify the future evaluation of the BCF information and to classify the different BCF topics in terms of their technical characteristics. In the first phase of the evaluation, the focus is primarily on semantic information, with the potential integration of other data and information sources already in prospect.