A standard quality vocabulary with specified semantics (in RDF) and a machine-readable representation (XML-RDF) together with the conceptual framework of the PICS environment will provide a common foundation for quality-based services.
There are several possible scenarios in which a quality vocabulary might be used. We now discuss a number of such scenarios in order to elucidate the requirements for the quality vocabulary and the technical framework in which it will operate.
It is useful to consider both scenarios for creation of ratings metadata and scenarios for their subsequent usage in the provision of quality-based services.
Third parties, such as subject gateways, may create rating descriptions and either make them available to users or develop services based on those descriptions. When the party providing a quality rating is not the same as the party providing the resource it becomes particularly important to ensure that rating information is current.
Information providers may provide ratings for their own materials. The provision of ratings that follow a standard quality vocabulary would allow evaluation and comparison of materials within a common framework.
Accuracy is a concern here. Clearly it is in an information providers interest to present favourable information about its resources. An information provider may wish to get quality statements endorsed by an official body. Quality statements certified with digital signatures may be a requirement.
On the other hand, a quality vocabulary may raise the information provider's awareness of quality issues. The process of self-rating may encourage the information provider to improve the quality of resources before making ratings available. The existence of third party ratings may have a similar effect.
Individual users may rate resources as they browse the Web. Their own ratings could, for example, be used to annotate their bookmarks. The use of a standard quality vocabulary would allow meaningful exchange of information between users and quality based services, even if that vocabulary was as simple as "I like this/ I don't like this".
A possible application of such data is the creation of a single collaborative rating value from the aggregation of a number of individual ratings (perhaps using mean or median rating values).
Within the context of this work quality based services form three major categories:
1. Filtering - does a resource meet a desired quality level?
2. Prioritisation - is one resource better than another (according to some criteria)?
3. Supplementary information - what information is available about the quality of this resource?
One possible quality based service is filtering out resources that do not meet certain pre-defined criteria. In terms of subject gateways or other third party services this filtering may be done at the resource selection stage. A quality vocabulary should provide support for filtering services including:
1. Subject gateways may accept only resources that meet certain criteria defined in terms of the quality vocabulary. The criteria guarantee a minimum level of quality for users of the subject gateway and for potential contributors. These may be concerned with subject or geographical coverage, educational level, availability of the resource, etc.
2. Browser-level filtering. For example, in an educational environment a browser could be configured to show only resources that meet certain criteria, this may be a statement about the intended audience of a resource.
Prioritisation is concerned with ordering resources according to statements concerning their quality. The 'suggest' and 'search' actions of the PICS Technology Inventory discussed earlier in the report are both forms of prioritisation of resources. In case of the 'suggest' action the context for the selected resource is set by an information provider (for example, the official Web site of a children's television programme may wish to recommend Web pages developed by fans). But the actual selection of the resource as opposed to other applicable resources, is made based on quality. For the search action the context is partially selected by the catalogue of the resources being searched and then narrowed further by the user's query terms; the results are then prioritised according to quality criteria.
A quality vocabulary should enable resources to be ranked according to (standard or user-defined) preferences.
1. Subject gateways may provide recommendations based on the quality of resources. For example, they may recommend the top five resources in a particular category.
2. Search engines may rank results according to quality criteria (as well as other accepted ranking criteria).
3. Users may be able to arrange their personal bookmark space according to quality statements related to the DESIRE quality vocabulary (either subjective statements made by the user or statements acquired from other sources).
It should be possible to use ratings developed in accordance with a quality vocabulary to provide additional information about resources. The 'inform' and 'warn' actions of the Technology Inventory are concerned with providing supplementary information. These actions correspond to two important scenarios for supplying supplementary information. The 'inform' action provides information once a resource has been accessed and the 'warn' action provides information before a resource is accessed. Supplementary information that is made available before a resource is accessed can be used to allow users to make an informed decision about whether to access a resource, e.g. providing a warning of the unsuitability of a resource for certain age groups.
Scenarios that make use of supplementary information include:
1. A browsing companion could show third party ratings associated with the current page and pages hyperlinked to from the current page (for example, by popping up a bubble when the focus moves over a link). Third party ratings would be obtained from one or more description bureaux. Ratings information may be displayed directly or processed with respect to the user's preferences (expressed according to a standard quality vocabulary).
2. An information provider might provide ratings as part of the description of its resources. The ratings may be collaboratively generated user ratings, `official' ratings of resources or self-ratings created by the information provider.
3. Similarly, subject gateways may provide ratings to users browsing a catalogue of resources. The ratings may be expert ratings provided by the subject gateway as a part of its service, or they may be collaboratively generated rating collected from subscribers.