Introduction

      NewsMaps are one of best examples of information mapping available on the Web today. They are also among the most map-like of information maps, borrowing literally and liberally from the cartographer's toolbox. The attractive and interactive NewsMaps maps provide a 'big picture' summary of large volumes of textual information represented as hills, valleys and white, snow capped, mountain peaks–a cartographic form common on topographic maps of the real world. They provide daily maps of international news, US news, and technology news [1]. NewsMaps was developed as a high profile 'show-and-tell' website to demonstrate the power of ThemeScape information analysis and mapping technology developed by Cartia, Inc. [2].

Screengrab of a typical NewsMaps map Courtesy of NewsMaps, Cartia Inc.

Mapping Information Mountain Peaks

      Developing graphical methods to represent the three-dimensional form of terrain on flat paper maps has long been a challenge in cartography [3]. Solutions developed for mapping relief include the use of contour lines and elevation shading. Contour lines trace areas of equal height and they are drawn at regular elevation intervals. To distinguish different contours they can be shaded with an appropriate gradation of colors that give the visual impression of changing elevation. This elevation color scheme is widely used and seems to make intuitive sense [4]. Both contour lines and elevation shading are successfully employed by NewsMaps to provide a terrain that appears almost three-dimensional on a flat computer screen. These maps are, of course, using the hills and valleys metaphorically to represent the volume of textual information.

      The spatial properties of location and elevation in the maps are used to encode key characteristics of the form and structure of the information space in a single visual snapshot. Firstly, the peaks in NewsMaps are formed by a large number of news stories discussing the same topic (which is labeled with keywords for the topic). Intuitively, the higher the mountain, the greater the number of news stories. The valleys are the natural transitions between one topic and another. The spatial concept of 'neighborhood' is also used, so the closer together two hills are on the map, the more similar their information content. Again, this is intuitive, as things that are closer together in space are usually more closely related. (Similar spatial concepts were used in the prototype information map ET-Map featured in February 2000 Map of the Month).

      To produce the daily NewsMaps maps the ThemeScape system applies sophisticated and proprietary lexical algorithms to analyze and in some senses 'understand' the content of text documents and the relations between them to distil the key topics and form the map. This complex process of turning huge volume of text information into a summary graphic representation like a map is called spatialization.

      NewsMaps are delivered over the Web in a fully interactive viewer, which allows you to explore the map on your desktop. By moving the cursor over the map's surface details on the individual news articles pop-up and in a couple of mouse clicks you can access the full story in separate browser window. You can also zoom in on a region of the map to see greater local detail or browse through a listing of the topics represented on the map or use a conventional key word search. The results of both of these techniques are clearly shown on the map, as the relevant articles are located with blue pin makers. One can also plant little red flags to mark interesting spots on the map and thereby identify the underlying article.

Cartia Inc. and Cartography

      NewsMaps is developed by Cartia, Inc., based in Bellevue, Washington State, a high-tech spin-off formed by information visualization researchers at Pacific Northwest National Laboratory (PNNL) in 1996 [5]. In an email interview for Map of the Month with David B. Lantrip, one of Cartia's cofounders, I asked him whether he thought maps were the right approach to the problems of searching and browsing the Web. His unequivocal answer was, “Yes! Or another yet-to-be seen visual interface that uses some spatial representation of the information such that it has many of the advantages we currently provide via maps.” He went to list these potentially powerful advantages, “... a high-level view of all the data at once showing patterns and relationships, visual recognition of document content similarity by spatial cues, scalability to millions of documents without any change to interface, point and click retrieval, attractive and visually simple display that encourages exploration, and various tools to explore map in detail.”

      Lantrip has a good grounding in spatial analysis, with degrees in engineering, planning and architecture, for his wide ranging role at Cartia developing advanced visualization techniques, evaluating how people use those systems and generally 'evangelizing the technology' of information mapping. Lantrip was part of the core team of researchers at PNNL in the early 1990s who were tasked by US government intelligence community to develop new visual interfaces to tackle the mounting problems of information overload. The researchers quickly hit upon cartography as a potentially powerful representational tool with a proven history, as Lantrip recounts, “we don't believe that it is an accident that records of maps predate evidence of written language.” Their prototype solution was called SPIRE - Spatial Paradigm for Information Retrieval and Exploration [6] and formed the groundwork for the development of ThemeScape. Designing information interfaces that can be used by the casual, untrained user perhaps ts he hardest part of developing NewsMaps, according to Lantrip. This is not dissimilar to real-world cartography, where designing a map that can be picked up and quickly and intuitively read by anyone is incredibly difficult [7].

      Development of ThemeScape is ongoing and one exciting future avenue is the use of their mapping as an alternative interface to the conventionally ranked list output of search engines. Lastly, I asked Lantrip what his dream map of the Net would be. His well considered response was for a super-charged NewsMaps style map capable of smoothly handling the whole Web, which would,

"...automatically updated every hour to include new pages. At the highest level, the map will show only very broad thematic categories as peaks with labels. Zooming into an area will recompute the map in milliseconds and show a new set of relationships based upon the themes in the view-closer the zoom, the greater the specificity of content. Searches will show hits plotted on the map and the user can recompute the map based on the query results only. Options will allow the user to constrain the map to exclude or include only certain themes/content and tools will allow the user to explore the map by specifying time and relations among people, places, organizations, and things. Map overlays will show user surfing history, other user behavior, theme growth rates, geographic information, etc. All this and the interface will be simple enough for a child to use!"

      Now this would certainly be a useful tool for exploring the topography of the Web.