Problems of Navigation in Document Space

3.5Navigation tools in Document Spaces

• 
  Overview diagram of the global information space and the local neighborhood of the current node.
  
• 
  Backtracking facility for going back to a previous page.
  
• 
  Interaction history including timestamps, footprints, and bread-crumbs. Timestamps record time and user movement and show when pages were visited. Footprints provide check marks in an overview diagram of visited pages. Bread-crumbs show check marks in an anchor of visited pages.
  

Gloor (1997) classifies navigation tools, related to hypermedia documents, into seven categories as follows:

• 
  Linking - links in hypertext. Links are also classified as static links or dynamic links.
  
• 
  Searching - full-text search engine such as WAIS.
  
• 
  Sequentialization - helping navigate by making a sequential path such as a guided tour.
  
• 
  Hierarchy - hierarchical display of hypertext structure in various forms.
  
• 
  Similarity - viewing based on document similarity.
  
• 
  Mapping - overview map of hyperdocuments.
  
• 
  Agents - artificial intelligence based techniques.
  

Based on function and information needed, the navigation tools are classified as follows:

• 
  locomotion provider
  
• 
  object information
  
• 
  position indicator, path and trail
  
• 
  overview diagram and map
  
• 
  search engine and filter.
  

Navigation tools normally provide many functions and combinations of information. These classifications try to capture the navigation frameworks and their functions. The movement in space is offered by a locomotion provider, a specific type of interaction with the interface. The information, given by object, position, path and local map, assists decision-making in the navigation process. Local map and global map explicitly show and help construct a cognitive map of the space.

3.5.1Locomotion Provider

In order to navigate through a document space, the interface must provide the ability to go to different places in the space. In a command based user interface, e.g. a file system, there is a notion of current position, i.e. current directory. Current location can be changed by giving a command such as “cd” to change directory. (It is interesting that a directory is considered a place but a file is not.) Files are objects in a directory. Thus, we may go from place to place and interact with objects in that place.
In current Windows, Icons, Menus, and Pointers interfaces (WIMP), a click on an icon may result in a new view, which is equivalent to changing places. This effect is strong when a new display replaces the old one in the same window. For example, clicking on a hypertext link brings up a new page. The target may be a directory, a hypertext link, or a control button such as “go back” in hypertext or “go up” in a sub-directory.
Another kind of moving in space is controlling the observer’s position in space. One example is continuous zooming, such as in Pad and Pad++ (Bederson & Hollon, 1994), providing visual cues about the egocentric view. In 3D presentation, users can move in the 3D space. This is a higher degree of freedom in movement than zooming and includes rotating one's viewpoint. In this view, users can be in a space between objects, in contrast with “click and go to,” where users jump from object to object. Moving is also done by controlling velocity. As in driving a car or spacecraft, a driver controls the speed of a vehicle and its direction.

An interaction is considered locomotion if it mimics a location change. Other interactions, such as manipulating a view based on observer rotation, may not count as locomotion. However, this action of manipulation is necessary for obtaining surrounding information used in navigation. While managing an exocentric view is not locomotion in the physical world, it is more difficult to say that it is not locomotion in a document space. From this view, locomotion may be extended to include a movement of a focal point.

Locomotion along a structure is different according to the structure’s shape. In a linear list, there are two directions from a current position. In a tree structure, there is only one way to go to the parent and many choices for visiting children. In a network structure, locomotion may be movement along any link. There is only one link that is immediately previously traversed but many choices in terms of where to go next. In a fully connected network, the alternative path is up to (n-1) choices, where n is a number of nodes. Every node is one link away from every other node.

3.5.2Objects and Object Information

The objects in a document space are documents or some abstraction of documents, e.g. collections of documents or parts of documents. The object is presented as a target and as an intermediate place in the navigation process. The object information, document attributes, and navigation records can be used in a decision process to select or not select each object.
In a command line interface, a presentation of documents is shown after the user issues a command. The object or place can be navigated to without being seen. However, the command usage data in a study by Kraut, Hanson, et al. (1983) show that “cd,” the change location command, and “ls,” the display objects-in- the-location command, which are the most and second most frequent used in the command line, are most likely to occur together, one directly preceding the other. In most GUI environments, objects or places are shown. One example of object presentation is that documents are presented as icons in a file manager system. The icons may be targets of a clicking action, which will open a new view on the documents. The icons might use different pictures according to file type. A list of files is another view that gives more information about each file than do icons alone.
The index, table of contents, anchor, bookmark, and history list may be used as active objects that allow jumping to some place. It is common to render each differently in order to indicate that they are clickable.

3.5.3Position Indicators, Paths and Trails

In order to show “Where am I?” navigation tools need to provide position indicators. These might consist of any of the following:

• 
  You are at the location x, such as a path name or URL.
  
• 
  You are at the document y, such as might be shown in the title bar of window.
  
• 
  You are in the collection z, such as a drive name.
  
• 
  You are here, such as a highlight showing the current selection state of a selected icon.
  

A path represents a traversal to a position that users can go to next. The path is different in “continuous go to” and “discrete jump.” In a “continuous go to,” the path is an interaction that allows users to move themselves in space, including between objects. For a “discrete jump,” the path is the same as the active objects from which users are allowed to select.

Position to go next may be specified as being absolute or relative. In the absolute mode, a destination place is directly specified in many ways including typing its name, clicking on its icon in an overview diagram or selecting a bookmark. The path may not be considered important in this case. In movement by relation to sense, a path is an available location to go to from a current position, mostly considered only as locally surrounding locations. This also includes moving backward and forward.
Landmarks are a significant place or environment. For instance, the most frequently used Web pages for each person may be considered an individual landmark in the Web site. Home pages and bookmarks are users’ explicitly defined landmarks in a cognitive map. Landmarks may be artificially added into space as aids for orientation, not as part of a data encoding scheme (Vinson, 1999). For example, backgrounds such as irregularly shaped mountains are added in virtual world. Tudor (1994) designed landmarks for GUIs and conducted research on search tasks. The result indicated that the addition of landmarks was significantly reduced search time.
While navigating an artificial space, the explicit presentation of a route is not likely. This is in contrast to navigation for traveling in the physical space where the path may be planned and marked on a map to indicate which way to go. Kommers, Ferreira, and Kwak (1998) presented a hypertext system that supports relation querying. Using two keywords, the from-to graph is generated and shown in between nodes and links.
Sometimes, path information may not be considered important. Tauscher & Greenberg (1997) analyzed Web usage logs and showed that the paths taken to a node have no strongly repeatable linear patterns. Bookmarks and “shortcuts” provide an alternate path for going directly to marked positions instead of navigating via the same path again. In a command-based interface, a location may be accessed by using a default search path. For example, to run a program or to request help, the program or man page file will be retrieved by searching a directory which is predefined in the PATH or MANPATH environment variables.
A trail, in our context, is a visited path. It may be used to answer the question, “Where have I been?” and “Where have I not been?” A trail is generated from interactions and paths. A trail is a record, with time stamps, of where the user has been. It is a list of visited places ordered by time. A trail may be presented to the user, combined with a document space information, in such a way as to show where the user has and has not been, what percentage of the trail has been covered, etc. When node traversal is recorded, it may record each visit to each node, with multiple timestamps for revisited nodes, or it may simply record the most recent visit to multiply visited nodes, allowing for a much smaller data set. Generally, trails consist of records of node visitations. While it is not common, particularly in the WWW, it is possible to construct trails of link traversals. This is of particular interest when there are multiple links between nodes, and thus the specific link traversed is of importance to understanding the trail. Trails may record only a short time span of navigation (within a single session) or a long time span (across multiple sessions). One implementation of a trail is generally called a history list.

3.5.4Overview Diagrams and Maps

Chen & Rada (1996) performed a meta-analysis which showed that a graphical map, a visualization of the organization of hypertext, is significantly useful in hypertext. Overview diagrams and maps provide an exocentric view of the space. They make sense of what the whole space looks like, how the space is organized, and how objects are related. There are many types of overview diagrams and maps of document spaces.
One class of overview diagram is a structure diagram. It shows an explicitly defined set of relations in a document space, such as the hierarchical structure of a file system or the network structure of hypertext. Another class of overview diagram is a semantic map (Lin, Soergel, et al., 1991; Fowler, Fowler, et al., 1991; Fowler, Kumar, et al., 1996). In this class, words in documents are processed. The process may not be truly semantic but rather an attempt to capture semantic aspects of the documents. For example, a semantic map may be created by projecting a set of documents into 2D or 3D and optimizing distance between them so that similar documents will be clustered. Document similarity may be measured by a distance vector method. Alternatively, similarity may be determined by some classification mapping. Diagrams of subject classification of books are also in included this class.
The overview and map are also classified as global or local views. The global view presents an overall view of space. It is relative to the size of pertinent space. For instance, a state map may be considered a global map if one is concerned with travel only within a city. Similarly, a map of all the files in a single disk map may be considered the global view even if there are many disks. The local overview/map view shows something around a local focus. It can indicate “where can we go next” from a top-view. A local view may be a “zooming in” of a global view.
A user may create an object that serves as a map. It is similar to a bookmark, trail, or history list, but different in that the envisioned document could have both locations in link form and content description. These may be viewed as a document or as guided tour. These objects, which may themselves be documents, can serve as navigation aids.

3.5.5Search Engines and Filters

Querying by submitting keywords, file names, and other metadata will result in a space of reduced size, which will then be more easily navigated. The search mechanisms and retrieval models are a matter of intensive research in information retrieval and will not be addressed here. Another approach to searching is to make some constraint on the space, for example, a dynamic query (Ahlberg & Shneiderman, 1994) which shows an overview of all data that would be returned by a query, and which allows incremental query modifications to filter out or to increase displayed data. More generally, a filter may be applied to a space before presenting data to users. For example, e-mail filtering systems will filter un-wanted emails from a given specification.
It is interesting to observe that filters are very difficult to implement in the navigation of physical space, but much more common in a document space. Thus, a document space may be constructed to show only a subset of the objects that would otherwise occupy the space. Similarly, while it is difficult to annotate a physical space, objects in a document space may be elaborately identified by pop-up cues, tags, or legends.
Many search engines, especially Web search engines, return large set of results (in the order of thousands). This large space needs some form of navigational tool. Search refinement processes, browsing approaches, or maps of results are some of the methods currently being used. The most common approach today is to order the retrieval set into ranked order and present them in small sets (10-25) at a time. One problem with these types of navigation tools is that the presentation has to be created upon the arrival of result, where a pre-computed presentation is not possible.

Download 324.2 Kb.

Share with your friends: