May 15, 2007 | In describing the benefits of the Semantic Web, people often say it is about open standards and data semantics. That is true, but it is too vague for most to see what they would gain from it. Specifically, it fails to satisfy those who want to know to what problems it can be applied, and what its advantages are. One needs to understand how to apply such tools to real-world informatics problems of data structure, discoverability, and usability.
To this end, a series of new W3C Recommendations are being developed in support of building and using the Semantic Web. What follows is a brief introduction to SPARQL, GRDDL, POWDER — the “Snap, Crackle, Pop!” of the Semantic Web.
SPARQL
SPARQL is a query language for RDF structured data. The equivalent of SQL for Web-data, it works in a similar way, using Select, Where, and From query parameters. The difference in SPARQL is that one constrains a query by using triples with variables (prefixed with ‘?’) rather than table fields, to match the RDF data:
PREFIX ls: http://lifesci.org/1.0
SELECT ?gene ?go_process
FROM myDataSource
WHERE
{ ?gene ls:has_Process ?go_process .
?go_process ls:associated_with <disease#Cardio
Vascular> }
When executed, the above query will find all genes in myDataSource with GO Processes that are associated with Cardiovascular Diseases. The information returned satisfies the WHERE conditional, and is a table of data when Select is called, and an RDF graph when Construct is used. SPARQL is meant to work not just with data stored in RDF (á la triples, see “The Missing Link,” Bio•IT World, March 2006), but also as an interface to existing relational databases (RDB); in such cases, SPARQL is dynamically translated to SQL calls, and the results are returned as RDF. This is especially useful when dealing with current data systems and legacy databases, and means that existent data can be readily ‘exposed’ as Semantic Web resources. By the time this article is published, SPARQL will likely be a Candidate Recommendation, a major step to becoming a Recommended standard.
GRDDL
Before one can query with SPARQL, there need to be RDF data sources. GRDDL (Gleaning Resource Descriptions from Dialects of Languages) is a mechanism that enables HTML and XML files to be transformed into RDF. Since most people have developed a lot of technology around HTML and XML, they are hesitant to throwing it away for something new. GRDDL allows the subtle insertion of a “hook” in such documents that allows those wishing to retrieve the file as RDF to perform a transformation on the HTML or XML document itself. Transformations are often performed using a specified XSL translator file referenced from within head tag in the original document:
<head profile=“http://www.w3.org/2003/g/data-view”>
<link rel=“transformation” href=“http://
www-sop.inria.fr/acacia/soft/RDFa2RDFXML.xsl”/>
<title>The Semantic Web</title>
</head>
The current file is used as the input of the XSLT transform and the newly generated RDF document is the output, which can be further processed and stored by servers. GRDDL will be a Proposed Recommendation in a few weeks time, also on the path of becoming a Recommended standard.
RDFa is an additional and related syntax for embedding RDF directly into HTML pages through the inclusion of tag attributes (property, rel) that say what the subject is, and its relations to the object resources or literals (strings):
<body>
<h1>A life science Semantic Web: are we there yet?</h1>
<dl about="http://www.doi.org/10.1126/
stke.2832005pe22”>
<dt>Title</dt>
<dd property="dc:title”>A life science Semantic Web: are we there yet?</dd>
<dt>Author</dt>
<dd rel="dc:creator” href="#a1”>
<span id="a1”>
<link rel="rdf:type” href="[foaf:
Person]” />
<span property="foaf:name”>Eric
Neumann</span>
see <a rel="foaf:homepage”
href="http://www.eneumann.org”>
homepage</a>
</span>
</dd>
</dl>
</body>
This is converted by the appropriate translator into the following RDF statements (i.e., subject-verb-object)…
http://www.doi.org/10.1126/stke.2832005pe22
dc:title “A life science Semantic Web: are we there yet?”
http://www.doi.org/10.1126/stke.2832005pe22
dc:creator http://www.myOrg.com/references/rdf_sem.html#a1
http://www.myOrg.com/references/rdf_sem.html#a
rdf:type foaf:Person
foaf:name “Eric Neumann”
foaf:homepage http://www.eneumann.org
Using RDFa, web pages can be interpreted by humans and software tools that can extract RDFa metadata directly from the page content. This metadata can be collected and stored in databases and used for further mining. It also allows plug-ins to find metadata such as meeting date and times, and push them into calendars. RDFa serves as a powerful, yet simple bridge between the current web and its future manifestation.
POWDER
Finally, we need to understand some of the ways RDF metadata should be used with existing resources such as web pages, scientific publications, and data. The POWDER Work Group, “Protocol for Web Description Resource,” will be developing a mechanism through which structured metadata (“Description Resources”) can be authenticated and applied to groups of Web-based resources. It will allow retrieval of the description resources without necessarily retrieving the full documents they describe, letting people and programs make decisions about whether they wish to retrieve or index the content based on its description. Metadata here includes authorship, dates, authenticity (trusting source), legal bindings, disclaimers, and other conditionals. These descriptors could have a major impact on how researchers efficiently find and access scientific papers and/or data. The resource metadata would be available both as part of general queries (e.g., find any documents produced by the NCBO project), and when deciding if the identified content has the appropriate label (e.g., for public consumption) or legal conditions (e.g., Creative Commons public-sharing).
For many, these recommendations and best practices will be the necessary pieces to enable them to begin building Semantic Web systems. The Semantic Web is moving from a vision to the establishment of foundational components that can be applied directly to a large set of informatics challenges. Researchers will be able to utilize these to organize and connect resource content and their meta-data according to reliable principles.
Eric K. Neumann is senior strategist at Teranode. E-mail: eneumann@teranode.com.
Subscribe to Bio-IT World magazine.