Selective Use of Full-Text Databases

Large files of natural-language text are now available for automatic processing in machine readable form. Such text files may include documents of textbook size, medium-size newspaper articles and short messages and mail items, and the subject matter may be effectively unrestricted. Typically, the stored material is not meant to be read sequentially from beginning to end. Instead, a selective, diagonal reading strategy may be preferred which skips among the text sections and paragraphs in accordance with individual user needs. Methods are described in this study for analyzing text files covering arbitrary subject matter, constructing links among text segments of varying size in accordance with computed similarities between texts and defining text traversal paths that are responsive to particular user needs. Such selective text traversal is useful in retrieving information from textbooks and instruction manuals and in consulting dictionaries, encyclopedias and other collections of text items. Topics: Hypertext construction, automatic text linking, selective text utilization, diagonal text traversal.