Feature

Information Models for Web Structure

by James Watson

All destinations on the World Wide Web are constructed from information. Yet, these destinations seem to be physical, and, as a result, must be structured to help users locate and navigate the information in an intuitive way. Imaginary maps called information models represent the varying ways in which information can be structured.

Cyberspace Defies Physical Laws

Though not subject to the laws that govern the physical world, cyberspace can seem to be a physical place. On the Internet, we can visit a company by navigating its Web site; jump to other Web sites; go backward and forward, or up and down; and return "home". The vocabulary of cyberspace is a manifestation of our desire to visualize it in the same way that we visualize physical space.

But cyberspace defies physical laws. We can instantly travel between two distant locations. Familiar routes do not lead to the same place every day. Unexplored paths cannot be reliably visualized.

Web Navigation

Web site designers must account for the differences between physical navigation and virtual navigation when structuring their little part of cyberspace. Neglecting important differences can leave visitors trapped in the Web like helpless insects.

Effective sites use conventional knowledge of layout to give varying importance to the items on a page. Text typically appears in a form that reflects the paper-based document from which it was derived. Elements that are unrelated to the text appear outside of its boundaries: at the top, left, and bottom of a page. (Technical limitations make it difficult to place items at the right of a page.) The most important items — for example, a link to the main menu — are best situated at the top left-hand edge of a page.

Within the pages beyond the home page, discovering where it is possible to go can be difficult. Few sites provide, on every page, information that enables users to situate themselves within the site.

Orienting the user is important in any on-line medium; but, on the Web, it is crucial. The designer cannot control the page at which a user will first enter the site. The user will not necessarily arrive at the home page.

A Web developer has as much responsibility to help those who arrive at an obscure page within a site to find their way around as to help those who arrive at the home page.

By basing the structure of the site on a familiar two-dimensional model such as those discussed below, a Web developer can provide users with the necessary sense of "virtual reality" as they navigate the site.

Information Models

The Web Model

One of the original visions for creating an on-line document was simply to convert some of its words into hypertext links. Each document in a collection could contain hypertext links to the others. The two-dimensional representation of this structure looks like a spider web:

Given the Standard Generalized Markup Language (SGML) origins of Hypertext Markup Language (HTML), it is understandable why documents were produced this way. A fundamental principle of SGML is that document parts, called elements, are identified and tagged according to their purpose; the elements can then be formatted differently for different media.

The now-familiar menus and button bars that appear in many Web documents were rarely used in these early documents. The words "click here to" never preceded a link, as they would have been out of place should the document be printed.

This "web model" has some flaws. Links in the narrative information interrupted the flow of the text. Users were swept away to another resource, sometimes with no way to return. And because hypertext links are highlighted, excessive linking could hurt the legibility of the text.

Moreover, users could not create a mental model of this information structure. While they might understand where they were, they could not easily determine what other information might be available.

Finally, because pages in the web model are not easily stored in a directory tree, and because the site structure is not related to the directory structure, the paths for the links are not intuitive. The links are therefore difficult to maintain.

The Tree Model

Typically, early Web authors and administrators were systems personnel who understood systems technology. The tree model established by some systems experts reflected the file-and-directory systems architecture. Users would navigate the tree in the same way that they navigated the directory structure in the underlying operating system. If users could remember the current directory and file, they would know their position within the site.

Some types of information lends itself very well to a tree structure. If the subject matter is already classified in this way, then setting up the Web site is simply a matter of replicating the structure.

However, people who don't normally work with a hierarchical file structure may not be comfortable with this paradigm, because a file is found in only one place within the hierarchy.

For example, a directory called "vehicles" might have one subdirectory for files about "cars" and one for files about "trucks". The "vehicles" menu would lead to two menus: one that lists files about "cars", and one that lists files about "trucks". Yet most documents do not have such clearly defined divisions. Where would a user find files about "cars" and "trucks"?

The Searchable Model

Advances in technology, hardware, and software have led to the searchable model. The user does not navigate within such a site directly. Instead, entry is to a page holding a form that, when completed, activates a search engine. The search engine reads a database that contains an index of files and returns a list of links to relevant pages. Many Web users today begin their "surfing" by first accessing a global search page such as Alta Vista or Lycos.

One of the main disadvantages of this model is that a search might return too many pages. Searching for even the most obscure term can result in a list containing thousands of documents.

Entire sites are built around search engines or related technology such as database access programs. These sites usually lack the hypertext links between files that facilitate navigation within a single subject or between related subjects. Some sites don't include files in HTML; they present downloadable files in a common word-processor format.

Pages are usually listed in an order that is based on the number of occurrences of the search term and the way in which the search term is used. Pages with the search term in the title or a heading are returned first. But, lacking the discretion of a human indexer, the list may not always be as useful as it could.

Nevertheless, the search-engine approach should be kept in mind. The designer should ensure that page titles are meaningful, clearly identifying the page and its content. Similarly, headings should accurately describe the information.

Paper-based Models

Books can be a foundation for effective on-line information models, simply because every literate person understands how books are organized and how to consult them.

Books can be classified by the way that they are used: as narratives, as references, and as textbooks.

A narrative work is read from front to back: for example, a novel. A novel does not usually have a table of contents or an index. A narrative is not typically opened in the middle (except when the reader is returning to a previous stopping point). The structure of a narrative is linear: you read the first chapter, then the next, and so on, until the end of the book.

The opposite of a narrative is a reference work: for example, a dictionary. Reference works are seldom read cover to cover. Rather, they are used to learn specific pieces of information. A dictionary is quickly scanned until the desired word is found, for example.

A textbook contains elements of both narratives and reference works, but has its own unique structure: for example, user manuals, school textbooks, cookbooks. A textbook has a table of contents and an index because, unlike a dictionary, its information usually cannot be organized into any obvious alphabetical or numerical order. The information can be accessed in two ways: it can be read from front to back, or it can be consulted to obtain a specific piece of information.

Hybrid Models

The models for most sites are hybrids: high-level access from the home page follows a tree model; documents are linked as in the Web model; and a search program is supplied so that users can access the site according to a searchable model. Documents (and sometimes menus) at the same level of the tree follow a narrative model.

By combining models, a designer ensures that users can access the site in the way that best suits their preferences. But hybrids require the designer to take care. When using multiple models, the designer must ensure that the model being used at a particular point within the site is communicated to the user. Otherwise, the user can easily become lost in a vast array of information.

Navigation Aids

Navigation aids expand the simplified models shown here. Many sites offer navigation aids such as a button bar or high-level menu on every page. These aids provide access from the current page to a set of other pages on the site.

The most common navigation aid is a Home Page button, that links to the page designated as the main point of access for a site: the main menu.

Less common is the Up button. The Up button links to the next highest level in the tree. If the user starts at the home page (main menu), then moves to a submenu, and then to a document, the document's Up button returns the user to the submenu. The document's Home Page button takes the user to the top of the tree: the home page. Should a user visit a site and arrive within a tree structure, the Up button can be useful in finding related information.

A Search button can call up a local search page, turning any site model into a searchable model. Every effort should be made to include a search feature on a site. Users then have two entry points to information: the site's primary model and the searchable model. Users can choose the model that they prefer.

Another button to consider is a Site Map button. A Site Map button links to a static page that shows a link to every page in the site. A site map is essentially a non-automated way to implement a searchable model. Tree models are easily illustrated on a site map, with files appearing in menu order. Indents can show the hierarchy. For model types, the site map could take the form of a topical or alphabetical list.

For paper-based models, Previous and Next buttons are very useful. These buttons differ from the browser's Back and Forward buttons, which navigate the browser's document history. Rather, in a multi-page collection of information, the Previous button goes to the previous page, and the Next button goes to the next page. For example, if you reach file #3 of a site from the main menu, the Previous button on that page would take you to file #2, whereas the browser's Back button would return you to the main menu.

Keeping the Balance

The first step in constructing a Web site must always be to identify and organize the information appropriately. The success of all further work and the site's ultimate usability depends on this step.

No single Web model lends itself to the organization of an entire site, each one can be useful in organizing selected documents on a particular subject.

By keeping many techniques in mind, by employing known models, and by communicating the models to the user, familiar and functional virtual spaces are created. Take care before rushing into a "unique" or "elegant" site design; it may only result in needless and confusing overhead for a user. At all times, balance the effort required of the developer or maintainer against the real benefit to the user.

This article is an excerpt from a presentation given by NIVA Inc. at the Society for Technical Communications mini-conference in Ottawa, 13 February 1998.