US Patent 7890859 Rapid similarity links computation for table of contents determination

In a method for identifying a table of contents in a document (8), an initial indexing text fragment at about the middle of an ordered sequence of indexing text fragments (14) is associated with an initial set of one or more candidate linked body text fragments selected from an ordered sequence of body text fragments (16). A lower-ordered indexing text fragment at lower order than the initial indexing text fragment is associated with a set of one or more candidate linked body text fragments at lower order than the highest order member of the initial set of one or more candidate linked body text fragments. A higher-ordered indexing text fragment at higher order than the initial indexing text fragment is associated with a set of one or more candidate linked body text fragments at higher order than the lowest order member of the initial set of one or more candidate linked body text fragments.