This paper deals with the CUMINCAD-repository (Cumulative Index on CAD), which was set up in 1998 and has served the CAAD-community since then as an important source of archived domain related information. CUMINCAD contains over 5,000 entries in the form of publications in the field of Computer Aided Architectural Design. The number has been growing steadily over the years. To date only advanced search mechanisms have been provided to access these works. This may work out well for a just-in-time location of a reference, but is inadequate for just in case browsing through the history of CAAD. For such applications, a hierarchical browsing interface, like one in Yahoo or is envisioned. This paper describes how the keyword categories were defined and how a moderate, distributed effort in defining the categories will allow machine-identified classification of the entire data set. The aim of the paper is to contribute to building up a wide spread consensus on what the appropriate keyword categories in CAAD are, and what sub-topics should sit below the main keyword categories.