This paper presents the RDFHomepage project, a framework for using a person's structured data sources to auto-generate an HTML homepage. RDFHomepage uses RDF files as input, and currently supports several well-known RDF schemas, such as FOAF. In addition to these we have RDF converters for other structured file-formats, like Bibtex. RDFHomepage produces valid HTML 4.01 Transitional pages, and makes it easy to roll-out functional homepages for a group of people. The generated HTML code is very general, allowing quick and easy page-redesigning using CSS. RDFHomepage is written in PHP and uses our system for generating PHP classes based on RDF class definitions, enabling quick and easy development of RDF handling PHP code.
RDFHomepage is a tool for automatic generation of HTML homepages based on RDF files and other structured information sources which a user might already create and maintain. Figure 1 shows an overview screenshot of the default homepage created based on a user’s RDF data. This page contains all things one expects on a typical homepage: the top shows the person’s name, email, telephone number etc., this is taken from his FOAF profile; further down there is a short bibliography, this taken from a file using the homepage-schema we created for this task; the next section lists the projects this user is involved in, taken from the DFKI Organisational Repository (OrgRep); finally the page has a list of people he knows, based on the friends he specified in his FOAF profile. RDFHomepage has the following attractive features: – Enables easy rollout of many identical websites across an organisation.
Section 2 outlines the architecture of RDFHomepage. Section 3 discussed the different datasources used by RDFHomepage and Section 4 presents our template engine for generating PHP Classes based on RDF Schemas. Section 5 outlines some future plans and makes concluding remarks. 2
For RDFHomepage we chose to use the web-scripting language PHP. We chose
PHP because it is free and open-source, and is a powerful and feature complete
language, with good support for RDF through the RDF API for PHP (RAP) [
The RDFHomepage distribution comes with a standard set of pages, including “About me”, “Projects”, “Interests” and “Publications”. The side-bar menu can be easily customised to include other static or dynamic pages, enabling RDFHomepage to integrate well with exisiting homepage components. There is also a selection of side-bar boxes available, showing elements like links to the raw RDF sources of the page, or links to companies, projects, etc.
Generation of each page compromising the RDFHomepage can take several seconds, depending on the size of the user’s data and of course on the web-server being used. A caching mechanism is therefore a part of RDFHomepage, which will only regenerate pages if the underlying RDF data has changed. Unfortunately normal HTTP/1.1 caching mechanisms could not be used for this, as a single page might be dependent on a number of RDF files, both local and remote, so the caching layer was implemented as an additional PHP layer around the page-generating code.
3.1
The Friend-of-a-Friend ontology [
BibTeX is a format for managing citations when using TeX or LaTeX. BibTex
defines different classes of publications, such as articles, books, theses, etc., and
associated optional and required properties of these. Most computer scientists
will keep a BibTeX file of their own publications up to date, for use when
selfciting or when publishing their papers on their website. There are many tools for
helping this process (for example JabRef5 and BibDesk6). Since people already
maintain this information in a structured format it makes sense to reuse this
information, and to this end we used BibTeX2RDF, written by Wolf Siberski[
This means users have an interest in keeping the names aligned, and may edit their BibTeX accordingly. We note however, that this is not really a technically satisfying solution. In addition to the author links, RDFHomepage will also link to the PDFs of any papers available for download, as well as generating links to the homepages of the venues where papers are published. The PDF downloads links are taken from BibTeX directly, URL being one of the optional fields of all BibTeX types, and most BibTeX tools make is easy to attach PDFs to citations this way. These links are retrieved from an internal DFKI wiki page listing conference abbreviations and corresponding web-links. To generate links to the corresponding conferences, a social approach was taken. This gathering of relevant conferences and their URLs was already being done in our research-group and the existing Wiki page is now simply parsed looking for HTML links tags with abbreviations as link texts. The conference names and URLs are extracted and matched to the names user in the BibTeX data. This motivated people to keep the Wiki page updated, having a similar effect as seen with the FOAF files. <aberer2003chatty:ACM Press> a bibtex:Organization; :ADR [ :Country “USA”; :Locality “New York” ]; :FN “ACM Press” . <aberer2003chatty:Aberer Karl> a bibtex:Person; :FN “Karl Aberer”; :N [ :Family “Aberer”; :Given “Karl” ] . (...) <aberer2003chatty> a bibtex:InProceedings; dc:creator [ a rdf:Seq; rdf: 1 <aberer2003chatty:Aberer Karl>; rdf: 2 <aberer2003chatty:Cudre-Mauroux Philippe>; rdf: 3 <aberer2003chatty:Hauswirth Manfred> ]; dc:date “2003-05”; dc:identifier “http://www2003.org/cdrom/papers/refereed/p471/471-aberer.html”; dc:publisher <aberer2003chatty:ACM Press>; dc:title “The Chatty Web: Emergent Semantics Through Gossiping”; dcterms:isPartOf [ a bibtex:Proceedings; dc:date “2003-05”; dc:publisher <aberer2003chatty:ACM Press>; dc:title “Proceedings of the 12th Intl. WWW Conference” ; :ADR [ :Country “Hungary”; :Locality “Budapest” ] ]; bibtex:pages “197-206” .
At the DFKI information about all projects and the people working in them is
centrally maintained in a Organisational Repository (OrgRep)7 [
kl.de/∼grimnes/2006/03/orgrep/orgrep.rdfs from http://www.dfki.unibeen mapped to getter-methods of the form get namespace localname, where the namespace is abbreviated. Each getter method takes a boolean argument specifying whether an array or a single value should be returned. Arrays are useful when a property is either multi-valued or the value is an RDF list or collection, if multiple values are present but only a single value requested a random one is returned. The values returned from these getters are either instances of other generated template classes, depending on the value being a typed resource and the appropriate class existing, or the raw RDF nodes otherwise. Note also that since RDFHomepage only does processing of RDF data, there are no setter-methods generated, however, these could easily be added following the same pattern. The template generator allows loading of multiple schemas before creating the classes, as is illustrated by the rdfhome htmltext property shown in the example, which originates from the homepage schema, but still has an rdf:domain of an OrgRep project. The template generator also supports inference, so for example: in the case of FOAF, one can generate a Person class which includes the properties from Agent (a super class of Person in the FOAF schema). The generates templates build on top of RAP and can seamlessly fit into a project already using the RAP API.
The approach of mapping RDF objects to a the Object Orientation
framework of a programming language is not new and is similar to many other projects,
for example: Tramp [
As shown by the large number of projects offering different solutions for mapping RDF data to classes in object-orientation frameworks, this is a very natural idea. As we see it there are two main reasons to use this approach: 1. Hiding the complexities of RDF – most programmers are comfortable with thinking in terms of objects and classes, whereas working with RDF graphs can be very complicated. 2. Enabling code-autocompletion – modern integrated development environments can be a tremendous help when writing object oriented code, and removes the need to know the ins and outs of the APIs one is using. With
} ?> <? class org Project { function org Project($uri,$model) function get rdfhomepage htmlText($a=false) function get org eMail($a=false) function get org homepage($a=false) function get org projectFlyer($a=false) function get org startYear($a=false) function get org endYear($a=false) function get org logo($a=false) function get org containsMember($a=false) function get org hasProfile($a=false) function get org informalDescription($a=false) function get org managedBy($a=false) function get org name($a=false) function getProperty($s,$p,$m=False) pre-generated code for RDF classes this convenience can also be used with RDF. 3. Adding custom code to wrapper – by creating classes that inherit from the generated classes one can embed functionality specific to particular classes in the most natural place. For example, foaf:Person could have an added function for automatically generating sha1 checksums of the email address.
With regard to the first point our RDF Templates were only partly successful, the users still have to understand how multiple values for a single property work in RDF, and still have to understand the difference between resources, literals and anonymous nodes. With regard to the second point, full code-completion is possible with our solution, and was used by several of the coders during devlopment. Code-completion being possible is another positive effect of pregenerating the classes, a downside of this pre-generation is that we are tied to a static schema, and cannot directly deal with properties that do not conform. For instance, if an instance of foaf:Person has a middleName property (which doesn’t appear in the FOAF specification), it cannot be accessed using a get method. However, a general getProperty method is generated to work around this. 5
RDFHomepage provides a quick and easy to way to create attractive and informative homepages, with all the content typically found on a research active user’s homepage. The pages are generated from RDF files allowing complete separation of content and appearance, as well as making the homepage machine processable by semantic web agents. Editing of a few simple configuration files makes customising the generated pages trivial, and if the customisation offered is not sufficient the use of auto-generated PHP classes makes writing additional PHP code to handle RDF very easy. The scenario of “creating a homepage” proved to be a good playground for Semantic Web technologies and as the commercial success of facebook.com and myspace.com show, it is a very active area. Deploying RDFHomepage at our department triggered users to keep their FOAF files, BibTeX files, the organisational repository and internal Wiki pages updated, and this in turn leads to more accurate date in other projects. We have several plans for further developments of RDFHomepage, primarily we continue to add sources of structured information that user’s may want to add to their homepage. Things we planning to add very soon include: – Flickr9 photo streams – for example, a side-bar box showing thumbnails of the user’s last photos. – RSS/Atom support – listing the last entries in the user’s blog. – Calendar support – where is the user today? What are his plans this week? This could for instance make use of the RDF Calendaring efforts10, which include converters between iCal and RDF.
There is also scope for improving our template system: to make it a more general solution for RDF based software engineering we need to generate “setters” as well as “getters” for the properties. Another important addition is to add support for data-types that exist in PHP, and automatically convert fields that represent for example dates to a standard format.
RDFHomepage is open source, release under a GNU Public License (GPL) and can be downloaded from http://rdfhomepage.opendfki.de. To get a better impression of what RDFHomepage can do, consider visiting any of the author’s homepages, they are all automatically-generated!11