Extracting variability information from software product descriptions is crucial when reverse engineering a software product line, e.g., for variability model synthesis. Existing methods are predominantly designed for feature-oriented product lines, where products are described by the set of distinguishable features they implement, and variability information may be expressed by logical relationships over these features. However, limits of such boolean feature-based variability modeling approaches have been highlighted, notably regarding their expressive power. In this chapter, we take a step towards more complex variability extraction and focus on extracting non-boolean variability relationships from multi-valued software descriptions. We first analyze software descriptions, variability relationships and extraction methods used in the boolean case. We attract attention to a knowledge engineering framework supporting a sound and complete feature-based variability relationship extraction method. The benefits of this framework include several extensions enabling to take into account more complex datasets than boolean ones. We explore one of these extensions to extend the traditional boolean extraction method and handle variability relationships including both boolean features and attribute values that can be used to synthesize extended variability models.