The Role of Meta-Data on the Dynamic Web

• Knowledge data published on the dynamic web may be held at many sites and included in many pages within these sites. For this reason it is necessary for all such data to be tagged with explicit reference to its context and source in addition to its triple of subject, predicate and object.

• For the subject, object, context and source the need for positive identification across the web is fully satisfied by citing URIs. The predicate must also be expressed as a URI but also needs an assurance that the associated definitions are unambiguous and meaningful to all comers. The only way this can be achieved is for the cited properties to be tied into a specific class within a single global ontology that covers the various perspectives of all users.

A Single Global Ontology

• This section outlines the key features that are required in the definition of properties cited as predicate in knowledge elements on the dynamic web. The best available standard for this role is RDFS but we must appreciate that this was developed without any specific objectives with regard to its use in web-wide operations . This issue is addressed further in the white paper I presented at STC06 entitled A Global Ontology for a Truly Dynamic Web.

• The global ontology as proposed here is designed to grow organically from an embryo comprising just the following two resources:
  • The concept of class with five meta-properties as follows:
    • name
    • descriptive comment
    • the class specializing scheme of which it is a member
    • properties applicable to instances
  • The class of all describable things. This comes equipped with one name and descriptive comment. This is the only class that is not member of a class specializing scheme. This is because it is the top of the tree. It has one property applicable to instances. This is called "name" and has a a special role in the searching for the URIs of things described by assertions. Other properties applicable to everything, such as "long name", "short name" may be added in due course. Expansion from this class of all describable things can be furthered by anyone by adding any number of subordinate class specializing schemes.

• From this simple foundation, the global ontology can grow without limit to include definitions of all classes of describable thing together with the properties (types of information) applicable to their instances.

• The global ontology allows for versions of properties so as to allow for minor differences of opinion on the precise definition of a property without losing any underlying broad agreement in concept. These subtle variations allow each assertion to be tagged with the exact meaning as understood by its creator at the time it was created.

Organic Growth

• The manager of any participating site must be able to augment this ontology at will through any of the following actions in respect of an existing class:
  • Add a new name - i.e. a synonym.
  • Add a new descriptive comment - this will be regarded by other users as a subjective qualification of the descriptive comment provided by the originator of the class and as a guide to the implications of any classifying assertions citing that class.
  • Add a new property - i.e. a new type of information by which instances of the class can be described.
  • Add a new version of an existing property - i.e. what the site manager believes to be a better definition for the same type of information.
  • Add a new subordinate class specializing scheme. This is the mechanism through which the definition of the initial class of every describable thing will be expanded into a single coherent ontology of all knowledge. In creating a new subordinate class specialializing scheme the manager of a participating site specifies and opens up a new basis of specialization which can then be used by itself and any other participating site to create new member sub-classes which inherit all the definitions and applicability of properties from the class (parent) for which that class specializing scheme is defined.

• The manager of any participating site must also be able to augment this ontology at will through any of the following actions in respect of an existing class specializing scheme :
  • Add a new descriptive comment - this will be regarded by other users as a subjective qualification of the descriptive comment provided by the originator of the class specializing scheme and as a guide to the implications of any member class created through that site.
  • Add a new member class. In creating a new member class for an existing class specialializing scheme the manager of a participating site expands the single global ontology with a class that reflects a particular perception which can then be embelished with applicable properties (types of information applicable to instances).

• The above facilities are all that is required. Nothing else is needed to enable the growth of the global ontology of the dynamic web to cover all classes of all things together with all properties through which they can be described. The magic of this approach is that it accomodates every single variant of perception while allowing every site to make full use of existing definitions by refining them to meet their individual needs. Examples of this organic growth will be provided in due course.

• Each property version is an individual semantic resource with its own URI. It is expressed by a single assertion whose subject is a property and whose value is the definition expressed in text.

Distribution

• This global ontology is a single coherent resource whose physical representation is scattered across the web. Each participating site must locate and acquire the definitions of all classes and properties needed to describe the information it uses. Further definitions are then added where necessary to form a locally stored subset of the one global ontology.

• The definitions that comprise the global ontology are part of the dynamic web and have the same format as other information. There are two ways in which a site can acquire copies of definitions relevant to its needs as follows:
  • By accepting offers of distribution from sites with similar interests - including sites whose prime purpose is to gather definitions from across the dynamic web and re-publish them to meet the needs of certain types of user.
  • By direct request to the originator at the locator identified in the URI of the property cited as the predicate of an assertion .