This report develops a new set of criteria for evaluating computer architectures that are to support systems with security requirements. Central to these criteria is the concept of a domain, here interpreted as a set of information and authorizations for the manipulation of that information in a comput system. Architectural requirements are grouped in three categories: logical structure, the physical structure. These criteria were developed in order to assess the utility of Navy standard computers as bases for secure embedded systems, but they are not specific to those computers.

A simple content-analysis program incorporated in a word-processing system can display the most significant sentence of a page of text and give a short list of the more important words. This could help authors write titles, summaries, and descriptor lists. The content-analysis program relies on word frequency, precedence, and co-occurrence as indicators of content significance. Test show it performs at least as well as some trained indexers.

Abstract Presented is an exploration of parallel processing as it might be applied to three functions fundamental to information science: sorting, searching, and evaluation of a search prescription.

Abstract Preparation of test corpora by computer simulation avoids the expense of converting real text to machine‐sensible format and forces the experimenter to explicate the basic assumptions underlying his keyword‐selection strategies. Its use permits unidimensional, quantitative, low‐variance measurement of the relative effectiveness of alternative selection criteria.

Abstract Appending a checksum (also known as checkring] digit[sl, or control digit[s] /sum) to a lengthy non‐mnemonic number sequence represents one technical control over transcription errors. Different checksum procedures are used by different nations (e.g., France, Denmark, Canada) for their personal identifiers (The U.S. does not use a checksum). This study reflects an attempt to discover the most effective procedure.

Abstract The conventional method of preparing SDI (Selective Dissemination of Information) profiles often necessitates detailed interview sessions between the client and one or more information scientists. This research shows that by submitting samples of the client's recent professional reading material to automatic text analysis, SDI profiles can be prepared that result in significantly higher initial recall scores than do those prepared by conventional techniques; relevance scores are not significantly different.