Document segmentation means dividing a large document into smaller pieces. This technique is used to reduce download time, so only the parts of a documents need to be downloaded that are actually of interest, not the whole document. Furthermore, especially with documents in HTML, having a document divided in small pages prevents that the user has to scroll a lot, and it's easier to bookmark a certain location than if that interesting place is buried somewhere in a long document.
Document segmentation in HTML is usually done one page for each (sub)section, resulting in many small pages that can be loaded quickly. Also, only the images referenced in that page are downloaded, not every one in the whole document. In sections only consisting of the section heading and no text, it's recommended to merge this heading with the next section (and it's text), so that no whole page is used for only a heading. Figure 5 illustrates this merging process.
With PDF, the story is a bit different, as the linking between different files isn't very well handled by the existing processing tools, and therefore rather big segments are recommended, usually containing a whole part or chapter of the document. If an index or working table of contents is essential for a PDF document, one might even considering not segmenting the document at all, to keep the link information.