The Purpose of this Document

• To enhance usability, by making explicit what assumptions were made about typical users of the space and therefore what user knowledge is required to use it effectively;
• To guide criticism, so that it reference both the design process and the design product;
• To stand independently as a way of reifying design knowledge for comparison, analysis, and reuse.

The Design Problem

The design of any information access environment must take into account a typical user's knowledge of the domain, his familiarity with search tools and search strategies, and the nature of the information accessible from the environment [2]. For the JAIR information space the typical user is assumed to be familiar with main topics of research within the field of AI and the terminology used to describe them. Also assumed is familiarity with navigating hypertext, using common user interface widgets (buttons, scrollbars), direct-manipulation interfaces, and full-text search interfaces.

The nature of the information accessed through the space is described in the journal's charter, excerpted here:

"JAIR's editorial board is dedicated to the rapid dissemination of important research results to the global AI community. The journal's scope encompasses all areas of Artificial Intelligence, including automated reasoning, cognitive modeling, knowledge representation, learning, natural language, neural networks, perception, and robotics."

Designing the Information Space

Dimensionality and Presentation

For this space, the primary space chosen to represent the documents was two-dimensional. Two motivations were the prevention of occlusion, which is a recurring problem in a number of information visualizations [3] [4] [5] and restricts focus at undesirable times, and the flexibility of leaving a third spatial dimension through projection to allow representation of additional information.

To make that dimension available, the two-dimensional layout is displayed in orthographic projection with a manipulable viewpoint. The projection permits information to be represented in the orthogonal dimension, and the orthographic projection preserves distance relationships throughout (as opposed to a perspective projection). The manipulable viewpoint allows the user to adjust the visualization to shift focus, prevent occlusion, and gives the impression of exploring a concrete artifact instead of viewing an image.

Article Layout

The categorization was performed manually. The topics were chosen from the 1991 ACM Classification System. 17 of the articles were given a primary category and a secondary category, when judged relevant to more than one topic. These secondary category assignments were used as a metric for category similarity; the more articles that were assigned to a pair of categories, the higher similarity value was assigned. Kruskal's multidimensional scaling algorithm [9] [10] was used to arrange the category centers to find a configuration in which the rank-ordering of dissimilarities is most closely preserved by inter-category distances. Kruskal's algorithm was modified to prevent category overlap. The desired interpretation of the category layout is that, on average, categories with more shared articles are closer together than categories with fewer shared articles.

Colors were chosen to emphasize the information-bearing elements of the visualization, and to render structural elements more subtly. This prevents visual distractions Tufte [11] calls "chartjunk," which hinder perception and interpretation without adding information content to the visualization. For example, the most important parts of the visualization for determining topic relevance and article access are bright green and yellow. Structural features, such as the category perimeters and the ground plane, are in dim gray.

Additional Features

• Search terms are kept in a per-session history list. The terms used in the last query are highlighted; terms can be added and removed from a query to quickly compose a new query by selecting and deselecting terms from the list.
• The result lines are broken into color segments, with each segment proportional to the number of occurrences of a term in the query. The line is scaled by the number of terms in the query. This permits comparison of results among successive queries, even if they use different numbers of terms. The segmentation also allows "visually boolean" operations, by visualizing the relative distribution of occurrences among all terms simultaneously.

Since downloading and viewing articles are expensive operations, any means of conveying more information about an article's contents beforehand is useful. The information space is augmented with a details-on-demand [13] feature that displays an article's full bibliographic entry and an excerpt of the abstract when the pointer is moved over the corresponding article-icon. This facilitates rapid browsing of the space's contents, an important feature in information access environments [14].

Two other browsing methods are supported: by author and by title. Full-list browsing, although inefficient from the standpoint of navigation [15], directly enumerates the complete set of articles and authors, which is not possible with fielded search.

Finally, when accessing an article, all of the documents associated with the article are presented as choices for downloading. In this way, the user can choose to view the article as HTML in the browser window, or download compressed or uncompressed PostScript versions. Each of these options has tradeoffs in download time and convenience for viewing, saving, or printing. In addition, many articles have additional files as appendices, which are accessible from the same list.

Platform Choice

Future Directions

References