Most database applications capture their data using graphical forms. Fields have limited size and predefined type. Although data in fields is associated with constraints, it is modeled to conform to a rigid schema. Unfortunately, too many constraints on data are not convenient as most human activities are document-centric. In fact, documents are a natural way for human information production and consumption. In this paper, we introduce DRUID, a document capturing and wrapping system. It ensures flexible and well-adapted information capture based on a Document User Interface and information retrieval based on databases. DRUID relies on a wrapper that transforms documents contents into relevant data. Also, it provides an expressive specification language for end-users to write domain-related extraction patterns. We validate our information system with a prototype.