In this position paper, we claim that useful information mined from social tagging systems can be published and shared as Linked Data, in order to be further exploited. The proposed approach aims: i) to be applied on any social tagging platform, and ii) utilize information that concerns the collective activity of users, not individual transactions. Such information involves topics of interests, emerging trends, events, communities around events/trends. To mine this knowledge from tagging data, a number of clustering algorithms are suggested.
With the advent of Web 2.0, tagging practices constitute a collective fashion of metadata creation, which is suggestive of individual users interests, since users tend to tag content that they ¯nd interesting. As more and more people have supported this surge, tag data provides a rich information source to study social patterns and emergent drifts/directions in the web user community. The main problem though regarding this information is that it is kept isolated in a number of proprietary systems and formats and, thus, it cannot be reused and further exploited.
Research carried out in the ¯eld of Semantic Web supports applications to
provide extensibility, °exibility, interoperability and reusability. To address such
issues in the tag space, many researchers claim that the application of mature
semantic web technologies (e.g. formal representation of content, formal
relations) on tagged data could add great value to the latter, as it may render a
kind of structure to them, remove ambiguities and promote re-usage of content.
A number of standards and initiatives in this direction involve SKOS [
In 2006, Tim Berners-Lee presented Design Notes on how to write Linked Data
on the Web, in which he outlined 4 basic principles that should be followed and
are summarized as follows: Each resource should have an HTTP URI, so that it
can be dereferenced either by users or by agents, as well as a semantic description
which should be represented in standard formats such as RDF/XML. Further,
the use of links that associate representations of the same resource in di®erent
datasets is encouraged, to promote discovery of related information regarding the
resource at hand. Following this article, a number of research initiatives appeared
that aimed at the production of linked data out of distributed datasets in the
web, the more widely known being the Linking Open Data (LOD) project1. In
October 2007, the datasets in LOD project consisted of over two billion RDF
triples, which were interlinked by over two million RDF links. By September
2010 this had grown to 25 billion RDF triples, interlinked by around 395 million
RDF links. Moreover, a number of applications were developed in which the
aforementioned principles are followed, as described in [
A detailed review on linked data initiatives and applications can be found
in [
We will now focus on what is currently missing from the Linked Data
landscape, concerning the interaction with datasets from the Web 2.0 landscape.
Given the popularity of web 2.0 platforms and the useful information that can
be extracted from these sites, as reported in the related literature([
App1 in Figure 1) { no mappings on data aggregation level (currently, the mappings are on individual data level, so the \Wisdom of the Crowds " is lost)
To address the aforementioned issues, we propose a clustering approach in the next section. 3
The current shared vision for the future is one of semantically-rich information and service oriented architecture for global information systems. As individual web 2.0 platforms are not embracing semantic tools to unambiguously de¯ne their data, this vision is still far away from reality. Here we present an approach for integrating web 2.0 data to linked data, and, thus, get structural information out of raw web 2.0 data. An asset of the presented approach is that it does not bring any burden on individual web 2.0 platforms; instead it can be applied on each one of them, providing that the platform o®ers access to the user data.
Furthermore, the proposed approach focuses on publishing as linked data not individual pieces of data, but information that is extracted from the collective user activity in a web 2.0 platform through clustering. The idea is given schematically in Figure 2 and summarized in the following steps: { Apply clustering techniques on a web 2.0 platform data, to extract interested patterns in the data { Map the terms mined in the extracted patterns to terms already in the LOD cloud { Publish the mined information as linked data and create RDF links to the mapped terms in the LOD cloud
Clustering approaches have been heavily researched for mining useful
information - such as topics, trends, user communities around speci¯c topics and
so on - out of tagging data in web 2.0 platforms. The clustering techniques we
propose to employ during the 1st step may be either i) co-clustering algorithms
to get groups of related content with associated tags, which can be mapped to
speci¯c topics ([
Having applied the clustering algorithm, the information that has been
extracted in each cluster should be mapped to terms already in the LOD cloud.
During this step, a semantic indexer like Sindice [
Finally, we propose publishing of this information as Linked Data, so that it can be further exploited. This involves representation of the data into RDF and creation of RDF links that connect the mined information with the mapped terms already in the LOC cloud, which were found in the previous step. 4
We envisage the employment of the proposed approach in the Future Internet, the so-called Internet of things, in a number of contexts that aim at employing jointly information coming from sensor data of \smart" devices and information coming from users' activities in web 2.0 sites. Publishing such information as Linked Data allows data that refer to the same topic, but come from di®erent web 2.0 sites to be interlinked, so that the maximum span of available information regarding a particular topic will be exploited.
Figure 3 illustrates a sample scenario of use of the proposed approach in the context of a smart city environment. Speci¯cally, we claim that the proposed approach can be used by automatic policy makers to improve a transportation system under special circumstances, or even to design a better mobility plan for the city. Smart cities will become a reality, when the objects around us become smart and begin to embody the functionality of our computers and mobile devices (Internet of Things). Thus, there will be internet-enabled cars, taxis, buses, cameras on the road and so-on. All such information will be forwarded to the policy maker agent to record the current status in the roads of the city. Additional information extracted from web 2.0 sites may signify events that are expected to cause disruption to normal city tra±c (e.g. concerts, protests). As a result, the moving vehicles will be aware of such events and act accordingly (e.g. in case of a concert with high popularity, a number of taxis may decide to go to that direction to serve the music fans).
In the long run, the sensor data coming from devices may be combined with analysis of information coming from web 2.0 users, so that new mobility plans can be proposed (e.g. bicycle lanes, pedestrian roads). The novel aspect here is that the proposals will be driven not only by the actual needs of the citizens as those are recorded in the sensor data of the devices, but also by citizens' opinions and emerging trends, as those are expressed through citizens' activities in web 2.0 sites (e.g. if a lot of people criticize the parking problem in an area, upload photos that illustrate the problem, a policy maker could evaluate parking de¯ciencies in the speci¯c area and recommend solutions such as additional parking, parking pricing and management strategies). The role of Linked Data in such scenarios is to link information that is mined by various web 2.0 sites that refer to the same topic. 5
Here we suggested a way to publish and consume linked data out of tagging data; the presented approach can be applied on tagging data of any web 2.0 application, provided that it o®ers API access to such data. The bene¯t of such a process is that dynamic information about a web community (such as trends, topics of interest of user groups, etc.) becomes available to all. This information is of particular importance to people/professions that are interested in listening the public pulse, such as businessmen and public administration bodies.