Automated text retrieval methods are used in libraries and in many online bibliographic and numerical data files. They provide the library user with fast and accurate access to books and journal papers. Text retrieval is also important in office automation systems where letters, reports, memos, and other documents are created, received, and filed for later retrieval. Office workers could benefit from automated information storage and retrieval systems. When these are combined with word processing and electronic mail, the amount of paper circulating in an office could be reduced. With the objective of applying text retrieval methods to an office environment, Faloutsos reviews various text retrieval techniques, indicates some of the difficulties involved, and describes some methods that he considers to be particularly applicable for use in office automation.

This is a well-organized and comprehensive paper which examines different text retrieval access methods: full test scanning, inverted lists of key words, multiattribute hashing (superimposed coding), signature files, and clustering techniques. Each of these methods has a long developmental history, and lengthy explanations would be required for a full description of the method with both its advantages and disadvantages. Faloutsos does an admirable job of summarizing, but unless the reader is aleady somewhat familiar with these techniques the exposition may be hard to follow. However, there is an excellent set of references, about 100 of them, which will enable the interested reader to pursue any topic in greater depth.

In the latter part of the paper, the author discusses the integration of text retrieval systems with database management systems appropriate for an office environment. He analyzes primary-key, secondary-key, and text retrieval access methods and compares these on the following criteria: space or memory utilization; response time for searching; handling of insertions, deletions, and updates; ease of growth; preservation of key-order; and the ability to integrate with other retrieval methods. The author concludes that, “In our opinion the signature file approach seems most promising for archiving documents in an office. It provides a reasonable compromise between the inversion method, which is fast on retrieval but expensive on insertion, and the full text screening method, which requires minimal insertion cost, but is slow.” But still, in the author’s words, “There are many open problems in the area.”