Modern Information Retrieval Chapter 10: User Interfaces and Visualization

Contents

3. Examples, Dialogs, and Wizards

starting points!examples example starting points starting points!retrieval by reformulation retrieval by reformulation

Another way to help users get started is to start them off with an example of interaction with the system. This technique is also known as retrieval by reformulation. An early version of this idea is embodied in the Rabbit system [#!williams84!#] which provides graphical representations of example database queries. A general framework for a query is shown to the user who then modifies it to construct a partially complete description of what they want. The system then shows an example of the kind of information available that matches this partial description. For instance, if a user searching a computer products database indicates an interest in disks, an example item is retrieved with its disk descriptors filled in. The user can use or modify the displayed descriptors, and iterate the procedure.=+1

The idea of retrieval by reformulation has been developed further and extended to the domains of user interface development [#!myers88b!#] and software engineering [#!redmiles91!#]. The Helgon system [#!fischer89!#] is a modern variant of this idea applied to bibliographic database information. In Helgon, users begin by navigating a hierarchy of topics from which they select structured examples, according to their interests. If a feature of an example is inappropriately set, the user can modify the feature to indicate how it would appear in the desired information. Unfortunately, in tests with users, the system was found to be problematic.
Users had problems with the organization of the hierarchy, and found that searching for a useful example by critiquing an existing one to be tedious. This result underscores an unfortunate difficulty with examples and dialogues: that of getting the user to the right starting dialogue or the right example strategy becomes a search problem in itself. (How to index prior examples is studied extensively in the case-based reasoning (CBR) literature [#!leake96!#,#!kolodner93!#].)

starting points!dialogues dialogues

A more dynamic variation on this theme is the interactive dialog. Dialog-based interfaces have been explored since the early days of information retrieval research, in an attempt to mimic the interaction provided by a human search intermediary (e.g., a reference librarian). Oddy did early work in the THOMAS system, which provided a question and answer session within a command-line-based interface [#!oddy77b!#]. More recently, Belkin et al. have defined quite elaborate dialog interaction models [#!belkin93!#] although these have not been assessed empirically to date.

The DLITE system interface [#!cousins97b!#] uses an animated focus-plus-context dialog as a way to acquaint users with standard sequences of operations within the system. Initially an outline of all of the steps of the dialog is shown as a list. The user can expand the explanation of any individual step by clicking on its description. The user can expand out the entire dialog to see what questions are coming next, and then collapse it again in order to focus on the current tactic.=1

starting points!wizards wizards

A more restricted form of dialog that has become widely used in commercial products is that of the wizard. This tool helps users in time-limited tasks, but does not attempt to overtly teach the processes required to complete the tasks. The wizard presents a step-by-step shortcut through the sequence of menu choices (or tactics) that a user would normally perform in order to get a job done, reducing user input to just a few choices with default settings [#!phelps94!#]. A recent study [#!carliner98!#] found wizards to be useful for goals that require many steps, for users who lack necessary domain knowledge (for example, a restaurant owner installing accounting software), and when steps must be completed in a fixed sequence (for example, a procedure for hiring personnel). Properties of successful wizards included allowing users to rerun a wizard and modify their previous work, showing an overview of the supported functions, and providing lucid descriptions and understandable outcomes for choices. Wizards were found not to be helpful when the interface did not solve a problem effectively (for example, a commercial wizard for setting up a desktop search index requests users to specify how large to make the index, but supplies no information about how to make this decision). Wizards were also found not to be helpful when the goal was to teach the user how to use the interface, and when the wizard was not user-tested. It maybe the case that information access is too variable a process for the use of wizards.

starting points!guided tours guided tours Walden paths

A guided tour leads a user through a sequence of navigational choices through hypertext links, presenting the nodes in a logical order for some goal. In a dynamic tour, only relevant nodes are displayed, as opposed to the static case where the author decides what is relevant before the users have even formulated their queries [#!guinan92!#]. A recent application is the Walden Paths project which enables teachers to define instructionally useful paths through pages found on the Web [#!furuta97!#]. This approach has not been commonly used to date for
information access but could be a promising direction for acquainting users with search strategies in large hyperlinked systems.