This report describes for the potential user a set of procedures for processing textual materials on-line. In this preliminary model an information analyst can scan through messages, reports, and other documents on a display scope and select relevant facts, which are processed linguistically and then stored in the computer in the form of logical content representations. To satisfy a specific requirement the analyst can initiate a search through the stored representations to identify those with relevant content. Once identified, the original texts from which those representations were derived can be recovered, and, with suitable on-line editing and the addition of commentary and interpretation by the analyst, formed directly into reports on the topics under consideration. The model SAFARI system has two phases, representation and recovery. In the representation phase, raw text is read by the analyst, who extracts the desired facts, phrases them into "kernel" sentences, and sumbits these "kernels" to the machine for processing and storage. The machine makes a semantic interpretation of the kernel based on its grammatical structure and formalizes this interpretation for storage. In the recovery phase, the machine displays a list of categories corresponding to the kinds of information it possesses, from which the analyst indicates the information he desires. (Author/RM)