Spreadsheets are heavily employed in administration, financial forecasting, education, and science because of their intuitive, flexible, and direct approach to computation. They represent examples for “active documents” which are considered very worthwhile in the field of Knowledge Management. But these activity traits also lead to usability and maintenance problems, as spreadsheet-based applications grow evermore complex and longlived. We argue that these difficulties come from a fundamental bias of spreadsheets towards computational aspects of the conceptual models underlying the applications, i.e., a (...) semantic bias. Concretely, we present the software architecture of the SACHS project geared towards a semantic enhancement of MS Excel spreadsheets that aims at compensating this semantic bias. (shrink)