Legacy enterprise systems mainly consist of two kinds of artefacts: source code and databases. Typically, the maintenance of those artefacts is carried out through re-engineering processes in isolated manners. However, for a more effective maintenance of the whole system both should be analysed and evolved jointly according to ADM (Architecture-Driven Modernization) approach. Thus, the ROI and the lifespan of the legacy system are expected to improve. In this sense, this paper proposes the schema elicitation technique for recovering the relational database schema that is minimally used by the source code. For this purpose, the technique analyses database queries embedded in the legacy source code in order to remove the dead parts of the database schema. Also, this proposal has been validated throughout a real-life case study.