From mobile phones, PDA to WebPad or even TV, the diversity of future web access devices will continue to increase. Current methods of Web content adaptation use style sheets as separate profiles that correspond to a document's appearance. However, only text-mode document-style re-formatting is supported by either of the two current standards, the Cascading Style Sheet (CSS) and the Extensible Stylesheet Language (XSL) [1]. As indicated in [2], more categories of style re-formatting are even in a stronger demand, e.g., mathematical characters with graphic-rich boxes, animated text/cartoon, graphic-rich advertisement, etc. Due to different capabilities of mobile terminals, it is even required to adapt web contents not only to different styles, but also to different modes.
A single-modal content reformatting is not sufficient to support the universal access of a web content. A multi-modal content adaptation represents not just the increase or decrease of display quality, e.g., font size and color. Even a text can have more than one mode of presentation, for instance, the mode of fixed text and the mode of animated text (a text with animation feature). Multi-modal content is usually easier to be identified when the content contains multimedia. For example, a video clip may contain raw data from video, audio in two or more languages, and closed text caption. Neither current W3C nor WAP standard is able to handle such content. In [3], a transcoding framework was proposed to adapt multimedia content into various types of terminals. In multimedia one can also think of transcoding the presentation axes of Spatial, Temporal, and Hyperlinking layout. On a large screen, images can be placed side a side, while at a small screen they may be better off in a time sequence. Scrolling is not always the optimal solution, for both the user interface and the content presentation.
Even a single text can have more than one mode of presentation. The following example illustrates a text in a multi-modal presentation.
Original content: Right Arrow
Normal text mode: <text>Right Arrow</text>>
Animated text mode: <text>Right Arrow <animate from="0px" to="10px" dur="1"/></text>
Pictogram mode: →
Graphic mode:
Animated graphic mode: 
Speech mode: <speak xml:lang="en-US"><para>Right Arrow</para></speak>
To support a multi-modal content adaptation, an architecture or a framework is necessary. However, it is not the purpose of this paper to propose a new architecture. Instead, we would like to address our interests on how content relates to users, i.e., the personalization. Sometimes We are using 'customization' to indicate adaptation to the machine, which is a kind of static adaptation, while 'personalization' is to the user, which is a dynamic process. In particular, our main interest on content adaptation is to ensure a proper content delivery according to
These are further explained in the following subsections.
Not all mobile terminals are made equal. Knowing a terminal's capability, the content can be properly adjusted for a better presentation. This is important especially when the variation of mobile terminals is expected to be very high. One can imagine that future personal information appliances can range from basic mobile phones with a black-and-white 4-line ASCII display, to more advanced PDAs with an SVGA 4-inch color display. Next to display, there is the user feedback interface. Mouse and keyboard can be suboptimal for smallscreen and mobile devices. Speech orientation appears natural on phones.
The terminal capability is not the only factor to determine the content adaptation. A user's preference can also be a very important aspect. Just like a personalized Web page, the content can be customized to each user. Currently, the efforts for Web accessibility to universal users have been conducted at W3C's WAI (Web Accessibility Initiative). However, it requires other optimizations such as the transcoding of a content to meet a user's preference or capability.
Due to the terminal mobility, it is also good to consider content adaptation based on the user location and/or context. For example, a user in a meeting may switch his handset from the speech mode to visual mode, in favor of silent content delivery. Technically, this can be handled at the composition layer using SMIL.
Today's Web technology is not sufficient for multi-modal content adaptation, although some of the issues can be solved or being taken care of. For example, there are issues on the need for interoperability to various terminal/user profiles, and the adaptation to layout a content with different presentation axes of Spatial, Temporal, and Hyperlinking. It is up to future discussions in related working groups whether we need a new standard or to make use of current methods and enhance them. However, as indicated in [4], the Web's continuing development brings more incompatibility problems to the browser. When achieving a multi-modal scenario, it is worth to spend a little effort such that the above mentioned problems are minimized.