next up previous contents
Next: INFORMATION SYSTEMS TEAM Up: RESEARCH ACTIVITIES AND Previous: Colorado Component of

INTERFACE DESIGN & IMPLEMENTATION TEAM

Membership: Nideffer (leader), Freitas, Lovette

Mission Statement of Team: The goal of the Interface Design and Implementation Team is to develop a web interface for the ADL that (1) represents the functionality and the content of ADL to users; (2) is responsive to the feedback obtained from ongoing user evaluation; (3) incorporates interface design concepts from ongoing research and development; and (4) advances the theory and practice of interface design for digital libraries and for geospatial search and retrieval.

Research activities over the last year
The Interface Design & Implementation (IDI) team developed a new Web interface to ADL, and a Java client and middleware that (1) represents the functionality and content of ADL to users; (2) was responsive to feedback obtained from users to the beta web interface; (3) incorporates user interface design concepts from ongoing research and development; and (4) advances the practice of interface design for digital libraries, and specifically for geospatial search and retrieval.

IDI began by reviewing documents provided by the User Interface Evaluation team: target user group activities; issue papers for search, retrieval and processing functionality; user scenarios; user requirement statements for search, retrieval, processing, interface and navigation, and content (what the library should contain). IDI also reviewed reports from the preliminary results of user evaluation studies on the prior Beta interface.

Interface Design Metaphors
Several initial designs for the interface were developed early on in the process to provide an overall conceptual framework regarding the information needs of the anticipated user populations (for example, earth scientists, GIS specialists, secondary and university students and educators, and library personnel). The need for several interacting components was determined: a map generator, a query area, and a results display and workspace.

It was decided that the map component should remain visible on the interface at all times; that search options should be presented in a small set of search types that hides the complexity of the metadata; that varying levels of search should be available for novice and advanced searching based on the task at hand rather than a category of user; that the query functionality should support iterative searching where the results of one search could be selectively used to form a new search; that the user should be informed of the total number of items in the result set, and be able to sort results according to temporal, spatial, and thematic attributes; that queries and result sets should be maintained during a session so that they are reusable; that the workspace should allow selected items from a result set to be stored.

Issues areas identified
Making map adjustment functions distinct from other uses of the map such as clicking on a footprint to get more information; designing protocols for the use of external tools, such as image-based searching, alternative map generation, manipulation of returned data objects, etc; enabling the user to understand the results of actions on various parts of the interface by utilizing an interaction flow/process orientation to design; implementing cooperation between the components of the interface; addressing scale issues related to displaying large numbers of footprints; designing consistent display formats for metadata; querying non-contiguous geographic regions and periodic temporal intervals; showing the user query progress indicators; flexible manipulation and reordering of result sets; creating a context map that reveals information about location, and lat/lon as the user navigates the main map; creating methods for displaying information about user selected map areas, and footprints; implementing varying degrees of help information.

Progress to date
The IDI team has made excellent progress meeting or exceeding the initial requirements list. In addition to successfully building a u usable client and middleware architecture to ADL, IDI presented their work at a variety of workshops and conferences, began to implement an external applications interface to allow ADL to hook into a variety of tools for querying, viewing, manipulating, and processing data. IDI also started research and development of tools for collaborative sharing and manipulation of multidimensional data in desktop computing environments (described in more detail below).

Previously Planned R& D Activities
RESEARCH TASK: Map & Query Interface
In the requirements list generated through discussion with the executive committee, the implementation team, and the user interface evaluation team, it was determined that the new UI would fulfill the following criteria: the system would present a unified search screen to the user (integrating the gazetteer, the catalog, and the map-based searching, which had formerly been kept separate); helping to reduce, and simplify the set of search types for metadata attributes (search buckets); allowing users to select search areas on map (rather than use whole map window as search area); allowing querying of non-contiguous spatial areas; provision of a summary view of the session to the user within the session. All of these requirements were either met or exceeded.

The new UI was completed implemented in the Java language as a standalone application. This allowed for much more control and freedom in design, while being able to address the issues mentioned above. Initially the entire interface (map, gazetteer and catalog, search, and workspace) were integrated into one window. This proved to be too confined, and the three major components were separated into floating frames which users could position freely on their desktop. The new map browser allowed for multiple non-contiguous areas to be defined, and in the same space, dynamic display of footprints from query ``hits''. The search window contained several panels of GUI widgets, which mapped to the ADL defined ``search buckets.'' Also, the catalog and gazetteer were integrated by treating them as equal collections, which allowed them to be searched simultaneously.

RESEARCH TASK: Workspace
A key strategy of ADL is to provide users with access to the information that is inferable from the items in the library collections. In particular, this implies that users must access varieties of items and apply procedures that extract appropriate information. The user workspace is where users can assemble documents by directly accessing them from the library collections, from parts of documents, or by applying information extracting procedures to documents. The initial design of the workspace contained little if any of this functionality, and was significantly enhanced.

The idea of user workspaces in which the processing of retrieved information can be performed was incorporated into the new user interface, as outlined in the requirements list. The Java client built by IDI would be designed to enable users to: stop a query and issue a new one; return to a previous point in the session; modify and reissue a query; view the total number of items in the results set; sort result set by search bucket parameters; display footprint distribution of result sets; set display formats for metadata; access low resolution browse images that can be enlarged for evaluation; allow selected items to be organized in a set of user-named collections; allow user-saved queries to be re-issued to the system in a later session; and allow storage of user-named geographic areas in a `personal gazetteer'. All of these requirements were either met or exceeded.

RESEARCH TASK: Help Services
The IDI team was solicited to build multiple levels of help services into the ADL client architecture. 7.2 Context-sensitive help. In the requirements list it was determined that IDI would create: an initial overview tutorial; user registration and login procedures to capture demographic data about ADL's user base, and assign unique session IDs; incorporate status of process indicators during queries; and provide mechanisms for ordering data that's unavailable online, or has access restrictions. All of these requirements were either met or exceeded.

RESEARCH TASK: External Tool Interface
The integration of data processing tools into the new UI was a high priority for the IDI team. Two pieces of software, an image-based querying tool developed by the Image Processing Team (led be B. S. Manjunath), and the Computational Modeling System, developed under Terry Smith, were ready for use with the ADL system.

A simple protocol was developed that allowed an ADL holding ID to be converted to the proper format and passed to any external tool. The architecture for adding tools is modular, so that they may be available separately from the UI, and either added or removed at any time. At runtime, the UI detects which tools are installed, and makes these available to the user for exportation and manipulation of items from their result sets. Also, a Java API was created to allow any group to ``ADL-enable'' their software, which is in line with the public interface for querying our middleware.

RESEARCH TASK: Spatially dense information representation
To communicate to users the extent and depth of coverage of a collection, the interface team created two types of graphical display. One type of display shows the spatial extent and depth of a collection; the other shows temporal extent and depth.

A spatial coverage display consists of a two-dimensional, color-coded intensity plot in which intensity corresponds to depth of coverage. Labeled axes and faintly drawn continental outlines give context to the plot without overly obscuring the background colors (which are the data of interest in this case). The color-coding is done using a perceptually linearized color scale that ranges throughout most of the visible spectrum; this allows substantial detail to be seen in the plot without giving visual bias to particular ranges of values. A legend adjacent to the plot correlates hue with numerical value. Below the plot is a table containing the numerical equivalent of the plot. We hope that the combination of the plot and the table will answer most users' questions regarding spatial coverage; the plot is good for seeing relative differences, while the table is good for seeing absolute values.

A temporal coverage display is similar to a spatial display. In place of the intensity plot is a histogram showing the depth of coverage as a function of time. The units of the histogram are years.

Coverage displays were created for the entire ADL Catalog and the entire ADL Gazetteer collections. This information was also refined by creating coverage displays for each (first-order) type of holding within each collection. Importantly, the scale and color-coding in the spatial coverage displays, and the scale and domain in the temporal coverage displays, are held uniform across all displays for a given collection, thereby making direct comparison of displays possible.

The coverage displays can be viewed by the user as HTML pages, and are accessible from the ADL interface. Automatic procedures were developed to periodically re-create the displays from the collection databases, thus keeping the information timely and accurate.

Another area of interest is in the presentation of large search results from a geographical database query. Problems include both the assignment of weightings to each returned footprint and also a mechanism for footprint clustering and the corresponding cluster displays. Finally, when results from queries of more than one database are being displayed simultaneously, as is the case with Alexandria, results from both the Gazetteer and Catalog must be visually tagged and easily identified as to their source of origin.

The current UI handles this by assigning unique colors to each collection available to the user, highlighting the currently selected footprint, and providing ``rollover'' information about footprints. The use of shading inside of footprints was also attempted, but proved to computational intensive for the current client.

A method of displaying an overview of the geographic density of the ADL holdings was implemented in the current ADL web interface. Areas with more items in the database are shaded a darker color than other areas. Several gradations of color are used to show various levels of geographic coverage. This visualization of the coverage of ADL holdings will be enhanced in the design of the new interface, where overviews on additional dimensions are planned: e.g. time and genre.

RESEARCH TASK: User Interface Evaluation Support
This is an ongoing task that will be undertaken in close co-operation with the User Interface Evaluation team (see report on user interface evaluation research).

At present, IDI has implemented procedures for allowing the middleware to manage a user database that records all of the user information submitted through the ADL User Registration page, where each user is assigned a unique identifier. This serves as the key for referencing all other user data, and allows UIE to do analysis while maintaining privacy of the user's personal information. As requested, the middleware provides an interface to creating and deleting user classes; keeps a session log that records all server side user actions (such as the start and end of a session) and user queries. What follows are detailed descriptions of the specific requests for data management from the UIE team, and to what degree those requests were met by IDI:

  1. Design separate registration pages for special needs (class assignments, evaluation groups, demos) so we can cluster sets of related users; we will be able to include or exclude any class of users when we view registration data; special registration pages will have a limited lifespan. This was accomplished.

  2. We will identify ADL staff as a 'class' of registrants and label them manually so that we can either include or exclude them from reports/analysis. We also plan to have a unique 'demo ID' assigned so that the logs from demonstrations can be can be clearly identified. This was partially accomplished.

  3. We will allow tracking of major announcements of JIGI availability (logging date, targeted user population, and version number) so that the effect of new versions can be evaluated in relation to data generated from evaluation of prior iterations. This was accomplished.

  4. We will allow tracking of registrations: date and number per day, week, month, cumulative total. This was accomplished.

  5. We will provide a method for determining the number of `active users' during the last two months. This was accomplished.

  6. We will build tools to generate reports for each data point on the registration form showing raw numbers and percentages of each choice. We want to allow UIE to extract individual or sub-group (ie, women, or a particular class) data from the aggregate data. This will allow UIE to look at individual or sub-group level data in relation to the aggregate. This was accomplished.

  7. We will work to establish the ability to generate correlations among data points on the ADL Registration page. This was accomplished, though formal analysis must be done through some other statistical analysis software.

  8. We will help develop ways for users to edit their registration information. This was not accomplished.

  9. We will allow UIE to bring back data regarding the number of times an individual logged on, filled out an exit poll, and sent an adl3help message. This was accomplished.

  10. The evaluative score data from the exit polls (system and content) will be cumulated and averaged, and the range and mode will be reported. We will also work to allow UIE to track these values for any particular person through time. This was not accomplished.

Abstracts of Published or Planned Papers

  1. ``Advanced Visualization Services: Collaborative Desktop Environments.'' (With Nathan Freitas). ``Individual Differences in Virtual Environments.'' Journal of the American Society for Information Science (JASIS). Edited by Chaomei Chen, Mary Czerwinski, Robert Macredie. Scheduled for late 1999 (vol. 50). In Progress.

    This paper addresses issues germane to: design and implementation of multi-user collaborative desktop virtual environments; spatialization of distributed hypermedia; data visualization metaphors; dynamic modeling of user behavior, alternative methodologies for search and navigation heuristics.

  2. ``Designing a User Interface for Access to Geospatial Information: Alexandria and Beyond.'' (With Nathan Freitas). ``Digital Images.'' Special Topic Issue of Computers and the Humanities. Edited by Abby A. Goodrum, Brian O'Connor, James Turner. In Progress.

    This paper addresses issues germane to: interface design and HCI; indexing, access and representation of geospatial information; facilitating scholarly exchange; designing to support multiple users and uses.



next up previous contents
Next: INFORMATION SYSTEMS TEAM Up: RESEARCH ACTIVITIES AND Previous: Colorado Component of



Terence R. Smith
Tue Jul 21 09:26:42 PDT 1998