WSEAS TRANSACTIONS on COMPUTERS Issue 12, Volume 4 ...

for new Web pages that are yet to be authored, it is
not practical for many existing Web pages.
We attempt to automatically generate a
multimodal mobile Web interface for the existing
Web pages without assuming any particular input as
in [19] or having the Semantic Web existing as in
SmartWeb [25]. In this paper, we focus on
addressing the Web forms component of Web pages
as part of our ongoing research.
3.2.3. Auto-Generation of user interface from
service description. With a focus on the human
facing services that have XML schema, a recent
research [22] introduced an approach for autogeneration
of user interface from service XML
schema. As we mentioned earlier, we have a current
effort [21] in generating multimodal interfaces for
mobile Web services from Web service description
(i.e. WSDL files) and this current work addresses
generating multimodal interface for Web pages. In
this research we also propose a unifying architecture
that facilitates generating multimodal interface for
Web contents including both Web pages and Web
services.
Web Page I Service Description Analysis
Extracting the content structure
Generating an abstract multimodal interface
Generating a concrete multimodal interface
Fig 3: The activity of generating multimodal
representation of existing Web page
or Service
4. Architecture
In this section we give an overview of the proposed
architecture and highlight the interface generation
process. We also discuss the architecture's main
components and the role of each component in
various activities.
4.1. Overview
As illustrated in figure. 3, the process of generating
multimodal representation of existing Web page has
four phases:
First, analyzing the Web page' structures. Web
pages, usually, have one or more of the following
components: pure text, graphics, hyperlinks, forms,
and embedded components such as java applets or
ActiveX components. These components present a
certain content structure through a designated layout
structure. Extensive analysis of these structures is
needed before adapting the Web page and
representing it using multiple modalities. Second,
extracting the content structure. Third, generating a
correspondent abstract multimodal Web interface.
Finally, generating a concrete multimodal Web
interface based on the generated abstract interface.
4.2. Architecture Components
The proposed architecture (see figure 4) has four
components, namely:
• Analyzer and Extractor
• Web Interface Generator
• Multimodal Proxy Server
• Multimodal Rendering Component.
These components can be deployed in a clientbased,
server-based, or proxy-based architecture.
However, we recommend a proxy-based architecture
where all components are situated at a proxy Web
server as illustrated in figure 4. Alternate
architectures are discussed in section 4.5.
4.2.1. Analyzer and Extractor. This component
receives Web pages or Web service description (i.e.
WSDL file) as input and applies a set of predefined
rules to extract content structure and produce an
abstract Web interface based on the extracted page
contents. As part of our ongoing research, a
prototype that considers the forms components of
the Web pages is under development.
4.2.2. Web Interface Generator. This component
takes the generated abstract XForms- based interface
created by the Analyzer and Extractor component as
the main input and also considers the device profile
and user profile in generating a concrete multimodal
X+V based interface.