Towel: Experiences of Augmenting the Web with Journey Knowledge Simon Harper, Carole Goble, Robert Stevens, and Yeliz Yesilada Information Management Group, Dept. of Computer Science, University of Manchester, Manchester UK simon.harper@man.ac.uk / http://www.man.ac.uk/towel 1. INTRODUCTION Movement, or mobility, is key to the accessibility, design, and usability of many websites. While some peripheral mobility issues have been addressed few have centered on the mobility problems of visually impaired users. We use our past work to address these is- sues and derive mobility heuristics from mobility models, use these heuristics to place mobility objects within a web page, and describe the construction of a prototype mobility instrument, in the form of a Netscape plug-in, to process these objects. Specifically, we likened web use to travelling in a virtual space, compared it to trav- elling in a physical space, and introduced the idea of mobility - the ease of travel - as opposed to travel opportunity. Our hypothesis is that travel and mobility within the web mirrors travel and mobility within real-world environments. We suggest that the Web commu- nity has typically concentrated on navigation and / or orientation rather than the whole travel experience, and that this neglect is cru- cial when dealing with browsing by visually impaired users. We therefore extend the definition of travel to mean: confident naviga- tion and orientation with purpose, ease and accuracy within an envi- ronment. Work, including ours, has shown that: Visually impaired users are hindered in their efforts to access the largest repository of electronic information in the world, namely the World Wide Web (WWW) [4]; A visually impaired user’s cognition, perception, and world view are highly egocentric, meaning that information feed- back should be tailored to these mental processes [1, 4]; Visually impaired web travellers are at a severe disadvantage, when moving around the web, compared to their sighted counterparts [4] (this is because of the lack of explicit and necessary mobility information, implicitly available to a sighted user [4]); The absence of suitable mobility guidelines, mobility design and evaluation methodologies, technical implementations, and work on holistic views (models) of mobility, all hinder visually impaired users [3, 4]; W3C WAI, We- bABLE, RNIB, AFB, and other guidelines focus on sensory trans- lation by the graceful conversion of visual to auditory information, but take no account of mobility; Hypertext design and evaluation methodologies, like HDM and SUE [2], pay little attention to mo- bility as do Web browsers for visually impaired users; Solutions exist to real world mobility problems of visually impaired trav- ellers. These solutions can be applied (after some conversion) to the web, so that a visually impaired user’s movement is enhanced [3]. We have now created a mobility tool in the form of a Netscape Plug-in in an attempt to support our theories. The plug-in pro- duces a series of fragmented web pages, decorated with generated information derived from the placement of the Towel Mobility Ex- tension (TME) objects. The plug-in is loaded when the browser Copyright is held by the author/owner(s). WWW2003, May 20–24, 2003, Budapest, Hungary. ACM xxx. encounters an XHTML EMBED element within the unprocessed hypertext file signalling that there are TME objects available. Once loaded the plug-in is driven by a series of keystrokes, which both activate and control its operation. These are function keys and are therefore all easily accessed by both visually impaired and sighted users. When the plug-in is activated the TME objects are processed and the page fragmented and saved as separate local hypertext files with an automatically generated preview page. This page preview is a clustering of hyperlinks to the individual page fragments. Each hyperlink is post-fixed with a word count of the target document so that a user knows what to expect when the fragment is reached. In this way orientation to the page contents is much quicker. In this paper we briefly summarise our evaluation of the tool and describe the lessons we have learned. 2. EVALUATION We conducted our experiment by first applying our craft based framework to seven sites representing a cross-section of the type of site available, from dynamic, through static, to document oriented locations. We analysed all sites from both a sighted and visually impaired users perspective. We then downloaded a page each from the sites and modified them in accordance with the results of the first application of our framework and re-evaluated the modified information. First, we focused on user evaluation which took the form of observing the responses of sighted and visually impaired users when moving around the processed ‘Towel’ test page. Users were asked a series of questions similar to those in the Pilot Study [4], and their responses where noted. We also asked more gen- eral questions regarding how the user ’felt’ about the reformulated pages, how easy they thought they where to move around in, and what improvements could be made. It was also explained that we where testing the ease of movement and not the visual style and so comments about presentation where excluded. It was found that the fragmentation of the page enhanced visually impaired users move- ment around, and cognition of, our test pages. Both groups said that they gained a greater understanding of both the page and site structure, and this was particularly apparent for visually impaired users. Small chunks of information worked best however some users commented that this would increase the number of clicks to get to some information on the page. The preview information en- abled the user to make a quick decision as to whether to investigate the page further or to move on to a different page. In fact visually impaired users found this far more useful than the sighted user, as these preview descriptions made visually implicit information ex- plicit. Users also suggested that more information about a link des- tination should be automatically included so that they didn’t have to directly investigate the target node. When investigating the page the journey metaphor of movement cues along the top of the document