Software product line engineering gathers a set of methods that rely on systematic reuse and mass customisation to reduce the development time and cost of a set of similar software systems. Boolean feature models are the de facto standard used to represent product line variability in terms of features, a feature being a distinguishable characteristic of one or several softwares. The extractive adoption of a product line from a set of individually developed softwares requires to extract variability information from a collection of software descriptions to model their variability. With the appearance of more and more complex software systems, software product line engineering faces new challenges including variability extraction and modelling. Extensions of boolean feature models, as multi-valued attributes or UML-like cardinalities have since been proposed to support variability modelling in complex product lines. In this paper, we propose research directions to address the issue of extracting more complex variability information, as a part of extended feature models synthesis from software descriptions. We consider the capabilities of Formal Concept Analysis, a mathematical framework for knowledge discovery, along with two of its extensions called Pattern Structures and Relational Concept Analysis, to answer this problematic. These frameworks bring theoretical foundations to complex variability extraction algorithms.