Search
Search

In the last instalment of my blog I introduced a Linked Data prototype application that exposes data of an SAP application as REST web service in semantic web standards so that it can be browsed with linked Data Browsers. My attempt was to link online documentation of an ABAP application server to external, official documentation. This kind of links can help SAP users to navigate through the jungle of different documentation types.

The most important aspect of this is that those links used for navigation are “typed” so that express a certain “meaning”. This can be a label that makes the property human readable. They can say that this is a link to normative and or perhaps non-normative documentation and much more.

Adding additional information, sometimes called semantics, is possible because the most important cornerstone are vocabularies defined by ontologies. Using them you can link two web resources using qualified links, called properties. These are far more than HTML hyperlinks that provide no further information about the kind of link. And last but not least: these metadata are readable by machines which can lead to much more sophisticated retrieval processes.

### The Interplay of Linked Data and Ontologies

If do you want to know how my prototype works you should have a look at the RDF data that is visualized using the Linked Data browser. I’m an XML guy so I hope you don’t mind the XML syntax. In fact there is a different syntax (N3, Turtle…) and it’s easy to switch from one format to another. Neverthelesse you should have at least basic knowledge about ontologies to understand the technical details of this weblog instalment.

In fact I used the DOAP Ontology that allows to express facts about the ECH project in NetWeaver Release 7.01. So please the XML elements doap:Project and doap:Version :

<?xml version="1.0" encoding="UTF-8"?><rdf:RDF xmlns="http://xmlns.com/foaf/0.1/"  xmlns:dc="http://purl.org/dc/elements/1.1/"  xmlns:doap="http://usefulinc.com/ns/doap#"  xmlns:abap="http://http://nsp:8920/zdocu/ontology#"  xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"  xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#"  xmlns:tab="http://www.w3.org/2007/ont/link#">  <doap:Project rdf:about="#ECH">    <doap:description>You use this component to control the system response in       the case of errors and conflicts in Service Operations. You can define wether       Service Operations are executed again after errors and conflicts or       terminated.</doap:description>    <doap:homepage       rdf:resource="http://wiki.sdn.sap.com/wiki/display/ABAPConn/Error+and+Conflict+Handler+%28ECH%29"/>    <doap:name>Error and Conflict Handler</doap:name>    <doap:programming-language>ABAP</doap:programming-language>    <doap:release rdf:resource="#NW7.01"/>    <doap:shortdesc>Error and Conflict Handler</doap:shortdesc>  </doap:Project>  <doap:Version rdf:about="#NW7.01">    <webpage       rdf:resource="http://help.sap.com/saphelp_crm70/helpdata/en/99/39092d836a4145b185cf574046c18a/frameset.htm"/>    <abap:containsReportDocu       rdf:resource="http://nsp:8920/zdocu/report/ECH_R_CUSTOMISING_HDS" />    <abap: containsReportDocu       rdf:resource="http://nsp:8920/zdocu/report/ECH_R_CUSTOMISING_HDS_COMP" />    <abap:containsReportDocu       rdf:resource="http://nsp:8920/zdocu/report/ECH_R_CUSTOMISING_HDS_COMP_SYS" />    <abap:containsReportDocu       rdf:resource="http://nsp:8920/zdocu/report/ECH_R_CUSTOMISING_HDS_SYS" />  </doap:Version></rdf:RDF>

So I used a common vocabulary for describing software projects. You can think of those ontologies as basis or foundational ontologies of the Semantic Web. But how does the Linked Data browser knows that vocabulary? This is easy: the app followed an URL. In fact I linked to an ontology that is defined using an XML-namespace xmlns:doap=”http://usefulinc.com/ns/doap#” and can be found at http://usefulinc.com/ns/doap# .

In the Linked Data approach I’m not only linking a resource to other resources – these links are qualified using an ontology. I used webpage and doap:homepage elements to link to external resources. And with abap:report elements I wanted to link to internal documentation of ABAP reports. Therefore I used a property that is used in another ontology:

xmlns:abap=”http://nsp:8920/zdocu/ontology#

In my prototype this ontology is exposed using a REST web service within AS ABAP in the following URL path http://nsp:8920/zdocu/ontology . In this ontology I defined property #containsReportDocu that realizes a link to an ABAP internal link documentation:

<rdf:RDF   xmlns="http://www.w3.org/2000/01/rdf-schema#"      xmlns:owl="http://www.w3.org/2002/07/owl#"    xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">  <owl:Ontology rdf:about="">    <description xmlns="http://purl.org/dc/elements/1.1/">    Ontology for SAP data</description>    <title xmlns="http://purl.org/dc/elements/1.1/">    Ontology for Linked Data</title>  </owl:Ontology>  <rdf:Property rdf:about="#containsReportDocu">    <comment>SE38 report</comment>    <isDefinedBy rdf:resource=""/>    <label>examples and test data</label>  </rdf:Property></rdf:RDF>

Of course the ontology can be visualized by a double click in a linked data browser:

In my opinion this AJAX-frontend is very useful and well-thought. The green light says that the described item can be resolved (which is trivial because it is part of the ontology). But in a linked data approach there are lots of references to resources in the internet and the link may be broken perhaps due to connection difficulties. In this case a red light would appear. If you click on the arrow symbol the linked object will be loaded in displayed in place. So a linked browser can display those information – perhaps even complex ontologies – in a way people can understand them more easily.

### From RDF/XML to Microformats

In my opinion this display possibility is great for the display of complex artefacts like  ontologies. But in other cases the strict separation between RDF/XML is cumbersome because you always have to provide multiple representations of the same resource. Therefore the W3C recommendation RDFa makes it possible to integrate RDF metadata into existing XHTML data and even into the intrinsic XHTML hyperlinks.

As a result you achieve metadata that is both readable and contains semantic metadata. This is exactly the same idea like Adobes XMP format that allows to insert metadata like geolocations, date, time and much more into existing binary-pictures.

In fact there exist more sophisticated Linked Data browsers like the Open Link Data Explorer Extension that has perspectives that show those metadata of an existing web page. Those browser sometimes use external services like http://www.uriburner.com/ to extract the semantic knowledge. At the moment I’m experimenting with my Linked Data prototype and those browsers.

### Summary

• Linked Data means emphasize to context of an infomation.
• Linking Data is more than just exposing knowledge as RDF – you have the possibility to link RDF vocabulary to ontologies.
• Ontologies are available on the internet so they can read on thy fly. This is a different use case compared to RDF and OWL datasets I used as data silos
• The RDF vocabulary can be easily extended by defining other own ontologies.
• One use case of ontologies is the definition of labels that are additional information that make Linked Data human readable.
• RDF and Ontologies are hard to understand. Fortunately there are powerful JavaScript libraries and browser add ons that make them easier to understand.
• With Microformats like RDFa you can embed RDF into existing XHTML documents.
To report this post you need to login first.

You must be Logged on to comment or reply to a post.

1. Chris Paine
It is always great to see an real worked example when something is explained – exactly what you have done!
Excellent stuff!
(0)
1. Tobias Trapp Post author
Hi Chris,

thanks for the feedback. The question why I used DOAP is in the most interesting aspect of blog. There are some basic or foundational ontologies on the we like SKOS or Dublin Core. My first reaction to them was that I would never use them. And later I realized that this changed and I’m using them very often: my Ontologies are annotated in Dublic Core for example and I’m only left that out because I already had problem getting a readable blog (see those linebreaks within the preformatted XML text).

My misunderstanding was that an ontology is something different than an XML schema and even an extensible XML schema. The differences are tha you can use only a part of an ontology and you can import other ontologies and extend you vocabulary and I’m starting to use this feature very often. The reason is that on hard problem of the Linked Data and the Semantic Web in general could be a variety of ontologies. On the other hand techniques that “translate” ontologies and enable an overall reasoning (think of \epsilon connections) are available in reasoners lieke Pellet but not standardized and not available in the Linked Data viewers I know so far. Perhaps this will change and Open Linked Data Server will support techniques like \epsilon connections.

So my approach was that I’m using everything useful I’m finding in foundational ontologies like DC or “second level” ontologies like DOAP so they form a basic language a lot of people and systems can understand. And extend them by own ontologies. In my prototype I exposed this ontology within the AS ABAP system but SAP could also standardize as own proprietary extension.

My summary is: IMHO a Semantic Web will need a translation mechanism like \epsilon Connections to deal with a variety of languages. But this doesn’t solve the problem because to many concept of the SAP world (or other proprietary worlds) are unique so there will be proprietary concepts in every case and Ontologies will help to define them in a standardized way. I already wrote a blog some weeks ago in which I suggested SAP should expose their eSOA methodology as ontology.

But there is a second option: SAP (and other IT vendors) could embrace more and more open standards. The could use eCommerce standards to structure their products and use more and more software standards like UWL. In this case a lot of concepts would be better comparable.

Cheers,
Tobias

(0)
1. Daniel Koller
Hi,

– re. Differentiation between Vocabulary/Ontology/etc. I found http://infogrid.org/wiki/Reference/PidcockArticle helpful. Basically a vocabulary means defined terms. (regardless of the technical format they are described in). In the semantic web sense an ontology has the value add to describe relations between different objects in a somehow domain-specific manner. (E.g. Martin Hepp from Munich developed GoodRelations, an ontology for eCommerce, which contains e.g. a Product, a Buyer, a quantity but also how they are related, more info at http://www.heppnetz.de/projects/goodrelations/ , schema pic at http://www.heppnetz.de/ontologies/goodrelations/goodrelations-UML.pdf )

– Dublin Core (DC) and Simple Online Knowledge Organisazation system (SKOS, http://www.w3.org/2004/02/skos/ ) are very good foundational ontolgies to start describing things in the semantic web, when you do not a specific vocabulary. Fried-of-a-fired (FOAF) and Semantically-Interlinked Online Communities (SIOC) are other very relevant foundation vocabularies.

– If Tobias found DOAP for something project-related, this is good, becasue it is much faster to start with “something” instead of trying to redefine “a project”. Writing up a vocabulary on the green field is actually hard work, as Richard Cyganiak noted down at http://richard.cyganiak.de/blog/2011/03/creating-an-rdf-vocabulary/ . If you later find DOAP to be not sufficient anymore, you can without any problems “mix in” further vocabularies, as there is no limitation to stick with one vocab for one resource(e.g.).

– If you come from a corporate viewpoint to define structures, you might want to extend a standard vocab and add company specific properties. (to avoid the maintenance effort of “nurturing your own baby”)

– schemapedia.com is a good starting point for retrieving defined vocabularies for different use cases: many entries contain examples how to apply it in the right way.

– Links between ontologies can be done using e.g. OWL constructs like e.g. the sameAS: this can then then be used in reasoning scenarios: currently.

– The fact that regular linked data viewers do regularly not show reasoning etc. IMO is related to the setup of the backend RDF stores. Reasoning/inferring new knowledge today takes place in the RDF store (and the results show up when e.g. querying it with SPARQL).

So far for now,

(0)