The importance of recording data of process and equipment malfunctions has been widely addressed and an enormous quantity of collections now exists. With problems regarding confidentiality being gradually solved and now with widespread use of computers, the need to develop computer-aided systems to help digest the large volumes of data is critical. Work so far has mainly concentrated on developing systems to help retrieval. Besides retrieval, the current approaches are not able to indicate further uses of the data. In this contribution, a prototype system is described that can be used interactively to discover causal knowledge from collections of text documents about process malfunctions and represent the knowledge in terms of rules and graphical models. This is a flexible system that gives users the freedom to control the knowledge discovery process, while at the same time providing users with maximum support of automatic functions. The system is also featured with a learning mechanism that allows continuous improvement of system performance during use. The graphical networks can also be used to retrieve the original text records from which a causal relationship is extracted.