![](bg3.png)
empowered with user-provided ontologies. The
addition of the ontologies is reflected as a new
canvas on the screen. Specifically, this canvas is
rendered at the lower-left hand side (see figure 3).
The ontology is depicted as a specialization tree.
This canvas is used as a rudimentary query
interface: the user clicks on the concept sought (e.g.
the RSS instance) and the canvas on the right shows
all news where this concept appears, regardless of
their container blogrolls.
This output requires a mapping between the
strings found on the news and the concepts of the
ontology. However, instances of the ontology are
identified through URIs, and these URIs will never
be found in the news. What is needed is a human-
readable version of the instance name.
To attain this aim, the notion of label is used.
RDFS provides a RDFS:LABEL as a human-
readable version of resource name. Analogously,
one wants to assign an instance with a human-
readable name even if it instantiates a class from a
given ontology that does not use the property
RDFS:LABEL per se. For instance, one might want
to state that the property siteOwner of the Site class
will serve to ascertain the appearance of a site in the
news so that the appearance of the string “Tim Bray”
in a feed will be taken as a reference to the resource
whose owner is Tim Bray. That is, the siteOwner
plays the role of the label for sites.
LABEL is then a kind of property. Data-based
properties (e.g. siteOwner) can then be of the special
kind label. Figure 1 depicts this through a tagged
value which is associated with the property. Notice
how a different label can be stated for each class:
stdName is the label for Standards whereas
siteOwner is the label for Site. It is also worth
mentioning that a class can have more than one
label, either defined among its own properties or
inherited from its superclass. In our running
example, all resources (should it be sites, standards
or tools) have as one of their labels the property
URL which is inherited from the resource class.
Hence, the appearance of either “Tim Bray” (the
owner) or “www.textuality.com” (the URL) will lead
to a connection with the very same resource in the
ontology.
For the domain at hand, basic word search seems
to be sufficient as both polysemy and synonymy are
rare, that is, “XML” or “Tim Bray” univocally
identified the standard concept, and the siteOwner of
the www.textuality.com site. This is due to the
smaller amount of documents to be tagged and their
more focused nature. Unlike other works where
tagging is extended to the whole web (Sheth et al.,
2002), this work focuses on tagging only those
feed’s items which have been previously selected by
the user. This implies that the context is much more
focused and hence, Tim Bray other than the owner
of the XML site would hardly appear in the feed’s
item.
Figure 3: Semantic query enhancement.
Being ontology-powered, RSSOwl queries can
now be stated at a higher level of abstraction. The
user clicks on any of the concepts that appear on the
lower-left hand side canvas, and the system displays
all news related with the selected concept. For
instance, it suffices a single click on Standards for
RSSOwl to retrieve all news which contain the labels
associated with the concepts Standards,
W3CStandards, RDF Schema, RSS and XML, which
would require several clicks otherwise (see figure 3).
Semantic view. Semantic metadata can also help
to provide a view more akin to the aggregator’s
perspective rather than feeds being clustered
according to their providers. Readers could thus
arrange feeds according to their personal
categorization schema, or being rendered along
indexing schema that fit the reader’s mental model.
For instance, if writing a report on XML, you
could be interested in highlighting the titles of those
items addressing XML issues, not to be overlooked
in the bulk of everyday news. In general, highlight
filters can be defined based on the presence/absence
of concept’s labels.
Another useful outcome is to annotate the content
rather than the title of the news. Figure 4 shows this
situation. An item has been selected. Its content is
annotated, i.e. the string “XML” is recognised as the
label of the XML concept. Rendering wise, this
leads this string to be highlighted, and turned into an
anchor to the URL of the associated ontological term
(e.g. http://www.w3.org/TR/1998/REC-xml-1998021
0). In this way, news items are enhanced with links
to the corresponding URLs, and in so doing, the user
POWERING RSS AGGREGATORS WITH ONTOLOGIES - A Case for the RSSOwl Aggregator
199