Next: Evaluation Plan Up: The Cheshire II Previous: Browsing and Relevance

The Cheshire II Client Interface

The development of a graphic user interface (GUI) for the Cheshire II system presented a variety of challenges. Some of these are common to the development of a GUI for any online catalog system; issues such as providing a simple, understandable mechanism for specifying boolean queries, making the full range of catalog functionality evident and easy to use,[5] and assisting the user with both the mechanics of the client's operation and the more general problems of search tactics and strategies are well-known to those doing GUI development for OPAC systems.[1]

The more difficult problems confronted by the Cheshire II system were those resulting from trying to provide a client capable of exploiting the full range of functionality presented by a variety of Z39.50 servers within a distributed network environment. For such a client to be completely successful, it must assist the user by providing:

  1. Aid in the initial selection of a Z39.50 resource appropriate to the user's query;

  2. Mechanisms for specifying a query that are adaptable to a variety of search engines (boolean, probabilistic, etc.);

  3. Support for display of documents which may vary by type (USMARC records, journal articles, etc.), language, delivery format (various kinds of text, graphics, etc.); and

  4. The ability to perform various post-retrieval operations on retrieved record sets, including printing, saving to file, e-mail, etc.

All of this must also be provided within a single, relatively coherent framework so that (as much as possible) the user is not forced to learn countless variations on the interface as it reconfigures itself to adapt to the current host. The above list is, of course, in addition to the normal requirements for a Z39.50-compliant user interface, such as allowing the user to make decisions in response to resource control messages from the server and providing intelligent (and, hopefully, intelligible) handling of the errors which may arise operating in a networked, client/server environment.

We have endeavored in designing the user interface to adhere to two principles: keep the client's functionality as visible as possible (in order to avoid having the user hunt through various menu layers for a needed command)[17] and minimize change to the interface when moving from server to server. See Fig. 3 and Fig. 4 for examples of how the boolean interface for Cheshire II reconfigures itself for searching on two different Z39.50 servers. Fig. 5 shows how the client interface reconfigures itself to enable probabilistic searching when connected to the Cheshire II server.

One distinctive feature of the Cheshire II graphical user interface is that it is based on the interpreted Tcl/Tk language[18], and can be easily modified to accomodate new interface features, including result summarization and reporting as found in the OASIS system[4]. The combination of Tcl/Tk and the workstation hardware being used for the evaluation experiments permits use of multimedia information sources including graphics and sound and will permit display of mathematical formulae and non-roman characters.

In addition to the Cheshire II client interface, the Cheshire II server will provide support for the HTTP protocol via an HTTP-to-Z39.50 gateway, giving access to popular WWW clients like Mosaic.

Additional Cheshire II Screens

This section doesn't appear in the printed version of this paper -- It provides access to some more recent Cheshire II screens.

The first screen image shows the initial state of the interface, before the connecting to any servers.

The next screen image shows the "Test Cheshire" database selected (The selection is made from a pop-up menu, which is not shown because of screen capture problems).

The next screen 3 image shows a query entered into the probabilistic search text box. Note that this query would fail to retrieve anything in a Boolean search.

The next screen 4 image shows the initial results of the query, displayed in the default "short" format.

The next screen 5 image shows the results of selecting the second record and then pressing the Feedback Search button. The prompter asks the user whether to perform feedback using all records retrieved or just the ones selected.

The next screen 6 image shows the results of the feedback search. Over 40 records having to do with extraterrestrial intelligence, life on other planets, etc occur at the beginning of the ranking, before any non-relevant items are encountered.

The next screen 7 image shows the Cheshire server reconfigured for Boolean searching (via the search type menu).

The next screen 8 image shows a MARC format display records from the Cheshire server.

The next screen 9 image shows the client connected via Z39.50 to the MELVYL UC Books database.

The next screen 10 image shows the client connected via Z39.50 to the Penn State online catalog database.

The next screen 11 image shows the results of a Cheshire search combining a probabilistic ranked search (geometry) with Boolean (author Chern).

The next screen 12 image shows the display for saved records. Any record seen can be selected and added to the saved records. These can then be printed, or mailed to the user's home account.

The next screen 13 image the "About Cheshire II display and logo.

Next: Evaluation Plan Up: The Cheshire II Previous: Browsing and Relevance

Contact: Ray R. Larson