Our research investigates how domain semantics are discovered from a relational database at a high level of automation. The discoveries are then represented as the extended Entity- Relationship schema  (i.e., conceptual schema) and integrity constraints of the database, which can support and drive the subsequent knowledge discovery in this database. This paper presents a discovery process which obtains an extended Entity-Relationship schema from a relational database by analyzing not only data instances but also the executable data schema of'the database. In addition, our research appears to be the first ones to address the problem of erroneous data for discovering inclusion dependencies. An interactive knowledge-based system, the Knowledge Extraction System (KES), has been developed to perform the discovery process. KES demonstrates how knowledge-based systems technology can be applied to support the work of knowledge discovery in a database. It also illustrates that the discovery process can be implemented at a high level of automation.