1   Discovery and Citation

The Concordia [1] project is applying principles of Web Architecture [2] and search engine methods to the study of the ancient world. Its principal objectives are:

1. To help information designers and developers discover external web resources that provide evidence supporting the authenticity or historicity of their own web resources.
2. To assist the same parties in discovering external web resources which cite their own web resources.
3. To do so at the least expense, using proven and readily available methods.

To Concordia, discovery is a feature that emerges naturally from the Web, and is assisted by conventional search engine methods.

Richard Cygniak has produced an evocative picture of linked datasets [3]:

Concordia aims to take the first steps toward creation of such a network for the study of the ancient world.

To date, the CIDOC Conceptual Reference Model (CRM) [4] has not played a role in creating an open, world-wide graph of ancient world data (see further, below). Concordia will defer implementation of the CRM and opt to catalyze the graph with a different, smaller, and not exclusionary vocabulary that better suits resources that are published on the web today.

2   Resources and Relationships

Information about the ancient world on the web today is communicated via web pages. These pages describe coins, inscriptions, persons, places, and other entities. They are either created directly by human authors or – in the case of many collections and reference works – are backed by relational databases or XML document collections. Such pages are linked consistently to other entities within their own sites or domains. There is little linking between entities of different domains, and what linking exists tends to be at a gross level. For example, the home page of a web site about ancient places might link to the home page of a site about inscriptions or coins, but linking between individual entities across sites is a nascent practice.

Concordia proposes to increase the linking of related resources across domains (e.g., coins to places to inscriptions to persons, all potentially on different sites). The resulting graph will represent a certain level of knowledge about the ancient world, knowledge that can be acted upon by software agents such as web crawlers or spiders, and indexed by search engines.

At a recent workshop [5], the following relationships were identified as prime candidates for linking:

• the find spot of an inscription or coin
• evidence for existence of a person or place in an inscription
• evidence for existence or characteristics of a person on a coin
• the place of birth or citizenship of an ancient person
• the place of manufacture of a coin

This is not the entire set of relationships among web resources representing or describing entities of the ancient world, but is a set sufficient to initiate a graph that encodes knowledge, promotes discovery of resources, and advances the field. The Web and Semantic Web [6] operate under the "open world" assumption, and so does Concordia. Discrete web resources must be identified via discrete URIs, but the initial relationships Concordia establishes between resources do not in any way preclude the existence of any other relationships.

3   Links and Vocabularies

On the non-semantic Web of HTML documents, relationships between resources are marked up using the <a> tag. Here is one common form:

origin: <a href="http://pleiades.stoa.org/places/639166">Xanthos</a>

A human can make use of this description – which asserts that the subject of the current page element (an inscription, in this case) originates from a place named "Xanthos" – but a machine can have no such sense. To a search engine, then, all such links say no more than that the web resource at the given URI (http://pleiades.stoa.org/places/639166) is "related" in some way to the subject of the current page element.

The primary Concordia relationships cannot be unambiguously expressed using the form above. For example, a machine cannot discriminate between the semantics of the following two lines unless it has been given specific information about the plain-text conventions used at each web site (a non-standard and non-scaling proposition):

origin: <a href="http://pleiades.stoa.org/places/639166">Xanthos</a>
attests to: <a href="http://pleiades.stoa.org/places/639166">Xanthos</a>

A small step toward disambiguating these is achieved by the following form:

<a href="http://pleiades.stoa.org/places/639166">origin (Xanthos)</a>
<a href="http://pleiades.stoa.org/places/639166">attests to (Xanthos)</a>

The strings "origin" and "attests to" within the <a> tags constitute a simple link type vocabulary. All that remains to make these sensible to a machine is to publish the types themselves on the web at permanent URIs, and use those URIs in a rel attribute:

<a
  rel="ov:origin"
  href="http://pleiades.stoa.org/places/639166"
  >
  Xanthos
</a>
<a
  rel="ov:attestsTo"
  href="http://pleiades.stoa.org/places/639166"
  >
  Xanthos
</a>

where "ov" is the vocabulary's namespace prefix. The syntax used in the markup above is called RDFa [7], a standard that will play a major role in the future of the Semantic Web.

4   On The CIDOC CRM

There are existing property or link relation vocabularies originating in the past-oriented digital humanities, the most notable being the CIDOC Conceptual Reference Model (CRM). The CRM is detailed, precise, and enjoys use in the design of museum databases. It is not, however, used openly on the web. Our assessment, which involved both web searches and an inquiry on the CIDOC CRM email list, has yielded no example of CRM properties being used to link ancient world data into an open, world-wide graph.

Considering now the inscription, coin, and place descriptions currently published on the web: the fine structure of the CRM has serious implications for the integration of these existing resources. Using the simple vocabulary of 2 terms shown in the previous section, we can immediately make a graph of resources as they exist now, without any remodelling, with no need to mint new resources:

If Concordia were to restrict itself to only the CRM properties, two problems would rise. New "event" resources would become necessary to mediate complicated relationships replacing the simple "ov:origin" property, and Concordia would lose the ability to indicate that one resource provides evidence for the existence or historicity of another. The situation is shown below:

What CRM property or combination of properties and new entities is to replace ov:attestsTo is not clear. Within CRM, it can be stated that a coin depicts a person, or that an inscription is about a person, but we are not comfortable with the assertion that these properties equal attestation. The ov:attestsTo relationship emphasizes evidentiary function in the context of historical analysis (e.g., a particular inscribed text attests to the use of a particular geographic name variant), whereas the depicts and is about relationships communicate just two of the conclusions that might be drawn from such analysis.

The new "finding event" or "minting event" resources (shown as red boxes in the diagram above) might one day be very useful in their own right, but Concordia needs to stick to practical rather than hypothetical objectives. Due to the short time-frame of the project, a requirement that new event resources be created for every origin relationship would stop the show. This is by far the greater of the two problems.

At any rate, the "open world" assumption of the web (and Concordia) does not rule out the future creation of event resources, nor the future use of CRM properties to relate resources initially bound together using Concordia's simpler vocabulary. In fact, it seems obvious that CRM will play an important future role in enriching a graph that is bootstrapped using a simpler, more generic vocabulary.