Profile-Based Adaptability in the Semantic Web

by Santtu Toivonen

One of the key challenges of the Semantic Web is the adaptation of Web content according to diverse parameters. In addition to differences between various devices, contextual information and preferences of users should also be taken into account in order for a flexible and proactive Semantic Web to emerge. Kontti-project, carried out at VTT Information Technology, aims to contribute to this vision.

In the Semantic Web, content should be adapted to many levels. Some elements that influence multi-level adaptability can be identified:

• Service providers: parties that produce Web content, eg yellow pages
• Resources: pieces of content, eg electronic weather reports
• Users: parties that consume the content
• Devices: terminals that are used when accessing the content
• Context: situational details to be taken into account when accessing the content.

Service providers are any parties that produce content in the Web, such as companies, the public sector, and individuals. Resources are individual pieces of Web content offered by service providers. Resources can be described with a number of attributes, such as content type, creator, creation date, price, etc.

Users in turn are any parties that consume the resources. The most evident users are human beings, but some higher level services can also be identified as content consumers. For example, consider a software agent collecting weather reports from several information providers in the Internet. This agent consumes the content and can thereby be functionally identified as a user. However, the same agent can act as a service provider to yet higher level consumers, such as a Web portal providing, among other things, information about weather. As far as human users are concerned, typical attributes to be taken into account when selecting and adapting content are age, sex, nationality, language, hobbies, etc.

Devices are used by human beings when accessing the content. The physical details of devices such as display size and amount of memory can have an impact on the adaptation. Consider again the weather report. When delivered to a small device with black-and-white character-based display, there is no sense in trying to express the report graphically. At VTT Information Technology, the content adaptation based on device profiles has already been studied in earlier projects. Contextual details such as time and place might also affect the adaptation; for example, a mobile user would typically prefer the weather reports of the location he is at or heading to. Kontti-project concentrates particularly on the adaptation based on these contextual details.

The adaptational relationships between the above-mentioned elements are depicted in Figure 1. Resources offered by service providers are adapted to the characteristics of different terminals. Contextual details and user preferences are, however, considered before device characteristics. In this way users receive the content they wish, in an optimal format and with situational details having been taken into account.

Adaptational relationships between the elements in the Semantic Web.

In the Semantic Web, the most natural way to express the characteristics of the different elements shown in the Figure is in the form of profiles. Users, service providers, and device manufacturers can create profiles in a language, such as RDF (Resource Description Framework), which is recommended by the Semantic Web community. Depending on the usage, some selected attributes from the profiles of various elements can be matched in order to deliver personalised material to the users. Kontti-project aims to set up a framework where users can design profiles for themselves and different applications, as well as other users being able to utilise these profiles.

The intention in the Semantic Web initiative is that in addition to human beings, computer programs can process Web content, for example to perform profile matching. In order to ensure this machine-processability and interoperability between different independent parties, the profiles should be semantically bound to ontologies. Ontologies, expressed for example in DAML+OIL (DARPA Agent Markup Language + Ontology Interchange Language) or its forthcoming successor OWL (Ontology Web Language), specify the meanings of the concepts and statements given in profiles. The weather forecast provider might create a profile for the report service using DAML-S - a language for describing Web services - on top of WSDL (Web Services Definition Language), and store it in a UDDI (Universal Description, Discovery and Integration) registry for other parties to exploit.

There already exist some standardisation efforts that at least partially cover the elements of the Figure. For example, for devices there exists CC/PP (Composite Capabilities/Preference Profiles), for resources Dublin Core, for services the above-mentioned DAML-S, for users vCard, for context data Wireless Village Presence Attributes, etc. While the ontologies must be extensible and specialisable, these standardisation efforts naturally cannot cover all the features people wish to utilise in their profiles. In Kontti-project, an ontology for expressing the properties required for constructing contextual profiles is created.

Please contact:
Santtu Toivonen
VTT Information Technology
Tel: +358 40 839 8405
E-mail: santtu.toivonen@vtt.fi