Making IrishGen SPARQL Part III: Ask and Describe

Christopher Guy Yocum

ORCID: 0000-0002-7241-3264

Over the last two posts (part I, part II), the two main query forms of SPARQL (Select and Construct) have been shown with examples from IrishGen. These two query forms together comprise the bulk of the operations a regular user of SPARQL will need to use on a regular basis. This post will explain the uses of the two less often used query forms: Ask and Describe. These will be covered in the order in which they are most often used.

Describe

The Describe query will return the RDF about a single URL supplied by the user. What is returned by a Triplestore is undefined by SPARQL (and deliberately so) in the standard. What is generally returned is the RDF that the Triplestore has about a particular URL. This is useful since, if the user knows the URL, they can interrogate the Triplestore about what is stored in the Triplestore about that URL without needing to go through the ceremony of a construct query. Moreover, Describe queries are often used in visual representations of RDF graphs.

For instance, the screenshot below is from GraphDB’s “Visual Graph” feature and uses Finn mac Cumaill as the exemplar. What happens is that the user sends the URL in which they are interested, GraphDB then makes a Describe query on the URL, and RDF returned is used by GraphDB to create a visual representation. This is very useful when investigating single individuals within IrishGen. Additionally, it is good if a user wants to have a secure, single starting point to explore the graph in various ways.

Using the “Expand” feature of GraphDB’s Visual Graph will send further Describe queries to the Triplestore. The resulting RDF will then be rendered in visual form and any connections will be presented to the user. Describe queries can also be done as plain SPARQL as well. For instance:

describe <http://example.com/LL/lagin.trig#Find>

will return over 1000 results giving everything that the Triplestore contains about Find mac Cumaill. The result of this is that the Visual Graph can be overwhelmed with information. The best practice here is to choose the predicates that the user is interested in rather than attempting to sift through all the information. While manipulating the Visual Graph feature of GraphDB is outside the scope of the present post, most often a user is only interested in one or two predicates so using the Visual Graph settings to filter out all results other than the ones that the user are interested in is generally desirable.

Ask

The Ask query has one use to inform the user if a basic graph pattern has a solution or not. No information other than a boolean that indicates existence or otherwise of a solution is returned.

In terms of IrishGen, this query form is used to ask for a quick indication if the information that a user is looking for exists within IrishGen. There is little point in using this in a SPARQL query when using IrishGen. A user will be better served by using a Select or Construct query as usually a user will need the results of a query rather than mere existence of a solution. However, as an example:

prefix irishRel: <http://example.com/earlyIrishRelationship.ttl#> 

ask
from named <http://example.com/LL>
where {
    graph ?g {
    	?a irishRel:nomName "Find"
    }
}

This query asks if there are any URIs that have the predicate irishRel:nomName and object literal “Find” or, more informally, if there is someone with the nominative name Find in LL. This will return YES as there is, in fact, a solution to the given basic graph pattern.

More broadly, the Ask query can be useful for what are called Federated Queries, which is an advanced SPARQL topic which will be covered in a future post. However, in terms of how IrishGen is currently configured, incoming Federated Queries are not possible due to the fact that the curators have not invested in server infrastructure that would allow interested parties to interrogate IrishGen through a SPARQL endpoint which is a URL that SPARQL queries can be send using HTTP without mandating that the user load IrishGen on their own machines. SPARQL endpoints are the standard method for ask and answering SPARQL queries across the Web.

Conclusion

This concludes the series on basic SPARQL in terms of IrishGen. While not every possible combination was explored, these posts provide a basis for further exploring the IrishGen database. Query languages like SPARQL, most famously SQL, form the backbone of data intensive systems. Learning to manipulate the four SPARQL query forms demonstrated in this series directly will give a user more power than any interface to a database could because an interface will always restrict the kinds of queries that can be run where writing queries in the language that is native to the Triplestore will allow a flexibility that a restrictive user interface will not.

From here, the user will need to use their new found SPARQL skills to explore IrishGen and other datasets. For instance, Wikipedia has two widely used datasets: DBPedia and Wikidata. Practising and using IrishGen will cement the core skills presented in this series and give the user a sense of the possibilities for further research. For instance, it is now possible to ask and answer statistical questions concerning the genealogies using SPARQL’s Aggregates or with genealogies from other sources, especially more narrative focused works, ask questions such as: were these characters connected to the secular genealogies and where and how? This can make the genealogies more accessible for studying narrative focused works and expand the analysis of those works beyond their immediate MS context. Questions such as the latter will become increasingly interesting as data is added in addition to the information contained in Corpus Genealogiarum Hiberniae.

Future posts will cover advanced SPARQL and RDF features such has as Federated Queries and Named Graphs. These future posts will facilitate advanced skills and deepen the user’s appreciation of the power of Linked Data and RDF.