Abstract: An information retrieval system processes a report-based data stream which includes report data having text lines comprised of fields, wherein each field is described by the type of data in the field. The system automatically abstracts the text line patterns from the report data by automatically classifying the text lines into species representative of text lines having a predetermined relationship to one another, automatically creating a definition of each species in terms of the species' constituent fields and automatically creating a virtual table definition specifying the hierarchical relationships among species based on functional type. The system automatically creates tables listing for each text line in the report, the species which the text line best matches, links entries in the list based on relationships specified in the virtual table, and then utilizes the linked list to generate virtual records in response to user-generated queries.