Evaluation of access methods to text documents in office systems

This paper compares two different approaches for indexing archived text documents. The first approach is based on inversion of words in the text, the second on the generation of a signature file representing the text content. A system reflecting the word inversion approach is compared against two systems reflecting the signature scanning approach and using, alternatively, superimposed coding and the concatenation of word signatures. Performances are estimated using analytical models of these systems. Characteristics are evaluated in function of office environment requirements. The evaluations derive from a model for estimating the statistical parameters of text archives.