A framework to support convenient annotation and intelligent querying of semantic Web resources. Annotation software is used by a Web site owner to generate RDF markup describing the content of their Web site. The RDF markup is essentially instance data that conforms to an OWL domain ontology. Web agent software crawls the Internet at regular intervals searching and extracting RDF marked up documents consistent with the domain ontology. The extracted RDF content is then stored in an RDF enabled database which forms part of a semantic middleware application, maintained on a Web server. The Web agents also have access to the domain ontology and store a local copy in the RDF database. The GUI is accessed remotely by an end user searching for Web content in the same way as a conventional search engine. The user requests are passed to the Web agents, which in turn formulate a query plan. Inferencing is performed on ontology schema information and instance data by the activation of a reasoner which is a component of the middleware. Queries are initiated in the middleware application by the agents, and results displayed to the end user via the GUI.

Web Site Annotation

An Annotated Website Generator

AcOntoWeb is a software application that allows accommodation Web site owners to automatically annotate their Web sites in accordance with a domain ontology. The software was designed by taking into consideration of the usability requirements that accommodation providers have for a new Internet technology, as ascertained by a survey conducted as part of the research (i.e. the software is easy to use, allows annotations to be added to existing Web sites, and will be made available as freeware).

They are then required to enter a description of the resort, choose a star rating and select a picture from file which will appear on the main page of their Web site. The next step is to select resort and room facilities that are offered by the business from a list of check boxes.

Finally the user is asked whether they would like to generate a new Web site or to annotate an existing Web site. If they choose to annotate an existing Web site, they are prompted to open the home page of their Web site and the RDF annotations will be imbedded and saved between the head tags. If the user chooses to generate a new Web site, a new annotated site will be created with link pages (E.g. Facilities, Location etc), based on selections made using AcOntoWeb. An example AcOntoWeb generated Web page and its associated RDF annotations are shown in Fig 4.

The RDF annotations start with namespace declarations. Namespaces act like a prefix to associate individual resources with a particular schema. They also define the markup languages used in the ontology. Conceptually, the annotations can be thought of as instance data of the domain ontology. Fig 4 shows the relationship between ontology schema information for accommodation and instances relating to the Sandridge Hotel. In the accommodation ontology, the class ‘Resort’ is related via an object property to the class ‘Location’, which in tern is related to the class attraction. The ‘Resort’ class is also related to the ‘Rating’ class and the ‘Facilities’ class which has 2 subclasses, ‘Room Facilities’ and ‘Resort Facilities’.

The ‘Sandridge Hotel’ is now an instance of the ‘Resort’ class and has the asserted location ‘Lorne’ which is an instance of the ‘Location’ class. It also has the asserted rating of ‘5 Star’ which is an instance of the ‘Rating’ class. It has an asserted room facility ‘Air-conditioning’ and asserted resort facilities ‘Bar’ and ‘Pool’.

Reasoning over Annotated Web Sites

A reasoner (also known as an inference engine) is an application that can read an ontology model with its schema and instance data, and derive additional knowledge about the domain. By applying a reasoner to the accommodation ontology extra facilities and attractions, other than those explicitly stated on the Web page, can be inferred as being associated with a particular resort. Fig 5 shows the asserted along with the inferred attractions and facilities of the Sandridge Hotel. Inferred entities are depicted in the diagram using a broken line.

With the use of class restrictions, the ontology shown in Fig 5 specifies that if a resort is rated 5 Star it must have a restaurant, pay TV and valet parking. It can therefore be inferred that because the Sandridge is a 5 Star resort, it must have these facilities. It can also be inferred by using class restrictions that because the hotel’s location is Lorne, and that Lorne has the attractions of surfing and hiking, that surfing and hiking must also associated with the Sandridge.