Editions   North America | Europe | Magazine


Xerox Scientist Finds Meaning in How Documents Are Presented

Press release from the issuing company

MILWAUKEE--Oct. 29, 2004-- When it comes to conveying meaning, how a document looks can be as critical as the words it carries, according to a Xerox Corporation researcher who is investigating document "intents" and the impact of cross-media design as a document moves between paper and digital displays. Whether a ballot is displayed on a touch-screen monitor or printed on paper, for example, its effectiveness depends on attributes such as design, the user's comfort with the technology, whether it is easy to find necessary information, and feedback about completion - values that are all quite independent of the content, said Steven J. Harrington, a research fellow in Xerox's Imaging and Services Technology Center in Webster, N.Y. However, the intent in either format is the same -- effectively facilitating action. Harrington worked out the concept of document intents and developed ways to quantify document properties such as aesthetics in collaboration with Rhys Price Jones and J. Fernando Naveda - from Rochester Institute of Technology - Nishant Thakkar, an RIT student now at IBM, and Paul Roetling, a consultant. They presented a paper today at the ACM Symposium on Document Engineering here entitled "Aesthetic Measures for Automated Document Layout." On October 22 at the American Association for Artificial Intelligence's symposium on Style and Meaning in Language, Art, Music and Design in Washington, D.C., they presented another paper entitled "On the Structure of Style Space for Documents." Harrington and his colleagues have identified more than 150 measurable value functions, including density of the text, colorfulness of images, regularity of positioning of images, and diversity of font and typeface, that a designer can use to convey the intent of a document. But their research explores the idea that although style is measured in a large number of value properties, those can be clustered into a relatively small number of intents. This finding has implications for the field of automatic document conversion. For instance, Harrington says that one intent of an advertisement is to capture attention. A printed advertisement might be designed to do that with a large headline; the ad on the Web might express the same intent with animation, or if it is displayed on a cellphone, it might include voice or sound. Because today's diversity of presentation methods makes any single fixed output appearance obsolete, Harrington envisions a transformation matrix. It would permit a document prepared for one type of display space to be analyzed so that its intents could be expressed accurately and automatically in the style space of another medium. With document layout a key means of expressing intent, the scientists considered rules for automating document layouts. They sought objective formulas that could be used to calculate the aesthetics of a document no matter where it appeared. Among the quantifiable factors they found that could produce aesthetically pleasing layouts were alignment, regularity, uniform separation, balance, proportion of white space, height to width proportion, uniformity and "page security" - the positioning of small objects so they don't appear to be falling off a page. "While application of these principles can yield pleasing results, we found that the combination of factors is not strictly additive," Harrington said. "In fact, one bad aesthetic feature can cause disproportionate harm to the overall appearance." Harrington and his colleagues conclude that there may be other factors influencing the aesthetics of documents, but the rules they have developed are relatively efficient and have yielded good results when applied to the problem of automatic document layout.