RDF Resource Search and Exploration with LinkZoo
Marios Meimaris
1,2
, Giorgos Alexiou
1,3
, Katerina Gkirtzou
1
, George Papastefanatos
1
and Theodore Dalamagas
1
1
Institute for the Management of Information Systems, Reseach Center ATHENA, Athens, Greece
2
Department of Computer Science and Biomedical Informatics, University of Thessaly, Volos, Greece
3
Department of Electrical and Computer Engineering, National Technical University of Athens, Athens, Greece
Keywords: Linked Data, RDF, Collaboration, Graph Search.
Abstract: The Linked Data paradigm is the most common practice for publishing, sharing and managing information
in the Data Web. Linkzoo is an IT infrastructure for collaborative publishing, annotating and sharing of
Data Web resources, and their publication as Linked Data. In this paper, we overview LinkZoo and its main
components, and we focus on the search facilities provided to retrieve and explore RDF resources. Two
search services are presented: (1) an interactive, two-step keyword search service, where live natural
language query suggestions are given to the user based on the input keywords and the resource types they
match within LinkZoo, and (2) a keyword search service for exploring remote SPARQL endpoints that
automatically generates a set of candidate SPARQL queries, i.e., SPARQL queries that try to capture user’s
information needs as expressed by the keywords used. Finally, we demonstrate the search functionalities
through a use case drawn from the life sciences domain.
1 INTRODUCTION
The Data Web has completely changed the way we
create, interlink and consume large volumes of
information. More and more corporate,
governmental and user-generated datasets break the
walls of traditional “private” management within
their production site, are published, and become
available for potential data consumers. The Data
Web extents current Web infrastructure to a global
data space containing and connecting data from very
diverse domains.
The Linked Data paradigm is the most common
practice for publishing, sharing and managing
information in the Data Web, and offers a new way
of data integration and interoperability. The main
concept in Linked Data is that all resources
published on the Web are uniquely identified by a
URI, and typed links (instead of traditional Web
hyperlinks) between URIs are used to semantically
connect resources. Reusing existing URIs rather than
creating new ones, and pointing from one dataset to
another by referencing these URIs, forms the Linked
Open Data cloud (Bizer et al., 2009).
Linked Data is mainly implemented with the
Resource Description Framework (RDF). An RDF
representation is a set of statements about resources,
known as triples, i.e. expressions of the form subject
predicate object. The subject refers to a resource to
be described. Actually, the subject is a URI
reference to that resource, which identifies it
unambiguously. Predicates are usually terms from
existing vocabularies and ontologies and are also
identified by URIs. Finally, the object can be either a
literal or a URI that refers to another RDF resource.
We will refer to triples whose objects are literals as
entity-to-attribute properties, and to triples whose
objects are entities as inter-entities properties. A set
of RDF triples can be represented by a directed
labelled graph, known as the RDF data graph.
However, in practice, RDF triples are stored in
relational database systems, native triple/quad stores
or graph DBMS (Faye et al., 2012; Bizer and
Schultz, 2008). To query Linked Data, the SPARQL
query language is used (Prud’Hommeaux and
Seaborne, 2008).
In this paper, we briefly describe LinkZoo
(Meimaris et al., 2014), a web-based platform for
collaborative management, editing and sharing of
Data Web resources, and we mainly focus on the
search facilities. Two LinkZoo search services are
presented: (1) an interactive, two-step keyword
232
Meimaris M., Alexiou G., Gkirtzou K., Papastefanatos G. and Dalamagas T..
RDF Resource Search and Exploration with LinkZoo.
DOI: 10.5220/0005499602320239
In Proceedings of 4th International Conference on Data Management Technologies and Applications (DATA-2015), pages 232-239
ISBN: 978-989-758-103-8
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
search service, where live natural language query
suggestions are given to the user based on the input
keywords and the resource types they match within
LinkZoo, and (2) a keyword search service for
exploring remote datasets, which automatically
generates a set of candidate SPARQL queries that
try to capture user’s information need as expressed
by the keywords used.
To demonstrate the services’ effectiveness, we
describe a real use case where the search facilities of
LinkZoo are combined to effectively address user
needs when working with a scientific Linked Data
set. Furthermore, we perform a preliminary
effectiveness study to evaluate the keyword search
service with query candidates. LinkZoo is available
at:
http://www.linkzoo.gr:9000
with
credentials (user: data_2015, password: data_2015)
for the demo account.
Figure 1: Linkzoo Architecture.
2 LinkZoo OVERVIEW
2.1 Architecture
The architecture of LinkZoo is shown in Figure 1.
The Storage Layer is built on top of a persistent
quad store, while the system’s functionality is based
on four basic modules: the Profile Management
module, the Resource Action Management module,
the View Management module, and the Search
module. The Profile Management module is
responsible for the administration of user accounts,
handling actions such as user administration,
account management, ascribing namespaces and
named graphs to users. The Resource Action
Management module provides processing and
editing of resources such as importing, annotating
and dereferencing. It is responsible for handling all
actions associated with each type of resource. Also,
it provides resource sharing functionality among
users. The View Management module controls the
lifecycle of views and folders; it also manages
containment relationships between resources, folders
and views. Views can be considered as different
workspaces where the same resources can be
organized in various ways. Finally, the Search &
Exploration Module is responsible for the searching
facilities implemented in LinkZoo. Specifically, it
implements different search mechanisms as well as
faceted browsing capabilities for private and public
user graphs. A more in-depth discussion of the
Search module is presented in Section 3.
2.2 Resource Model
The LinkZoo Resource Representation Model
captures the following aspects: (i) resource
descriptive metadata, (ii) resource interlinking, and
(iii) view definitions and containment relationships
of resources in views and folders. Common
vocabularies such as RDFS, Dublin Core Terms and
FOAF are used to model non-functional metadata
(e.g. labels, creators, etc.). Moreover, users can
import existing ontologies or define new ones under
their own schema namespace. Given that a resource
can co-exist in many user accounts (e.g., in case two
users happened to import the same resource),
resource definitions and views in LinkZoo depend
on user context. Multiple parallel versions of
resource definitions are stored in their owners’
named graphs. LinkZoo handles a variety of
resource types, e.g., files, URLs, contacts, RDF
datasets and remote SPARQL endpoints.
Furthermore, folders, i.e., resource collections, are
also modelled as a special resource type, and, thus,
can be annotated, shared and linked accordingly. We
have defined an extensible taxonomy of resource
types that includes various levels of specialization
for each type. This way, we allow for different
handling of each resource type or sub-type.
2.3 Resource Annotations
In LinkZoo, users can annotate resources and enrich
their definition with new triples. Many established
ontologies and vocabularies have been imported in
the tool for quick access, while new properties can
also be created on demand, under each user’s custom
schema namespace. Annotation can be performed
manually and collaboratively, as well as
RDFResourceSearchandExplorationwithLinkZoo
233
automatically. In the case of URL resources,
external APIs are used to automatically enrich the
imported URLs by parsing their content and
extracting related Linked Data entities (specifically
DBPedia and Freebase resources). In particular,
LinkZoo utilizes the Alchemy API (Alchemy API,
2015) but other similar services can be used as well.
2.4 Sharing and Collaboration
LinkZoo resources can be collaboratively annotated
and enriched with new knowledge. This is achieved
by sharing resources with other users, with
appropriate roles and privileges. Three levels of
privileges, represented by three user roles, ensure
proper sharing and usage among users. These are the
owner, editor and viewer. Owners and editors of
resources are able to share/unshare, annotate and
delete them. Viewers cannot perform any kind of
write-related operation that alters the state of the
resource in the storage, and are thus limited to read-
only actions of their shared resources. Furthermore,
resources can be private or public. Shared
directories pass on their sharing status to their
contained items, and whenever a new resource is
inserted into a folder, it automatically becomes
available to the folder’s shared users.
2.5 Linked Data Publication
Creating and publishing resources as Linked Data is
a key feature of LinkZoo. This means that created
resources are automatically assigned dereferenceable
URIs, which can be used for external linking and
referencing. These URIs follow a simple minting
schema that takes into account the type of resource
as well as a unique identifier created dynamically.
Dereferencing is performed when there are
appropriate permissions, thus restricting external
users with no authorization from getting access to
descriptions of private resources. Unauthorized
dereferencing returns a limited description.
However, for a private resource, a public
dereferenceable URI can be generated on demand,
allowing the user to offer access to others without
changing its status. If the user owns a LinkZoo
account, he can choose to import the item to his
account. Also, the platform offers serialized RDF
exporting facilities for selected resources.
2.6 Static and Dynamic Views
The default exploring and browsing mode of
LinkZoo follows the traditional folder-based
approach of file systems with visual interfaces.
However, LinkZoo exploits the semantics of the
resources to provide multiple ways of organization.
Users are able to organize their resources based on
their properties and store the results as linked views.
Views leverage the semantic web by offering
intuitive means for organizing, searching and
discovering new resources either within the platform
or the entire LOD cloud. In essence, views act as
workspaces and can be specialized in two sub-types,
namely static and dynamic. These can be
parallelized with materialized and non-materialized
views in relational models respectively. Dynamic
views are result sets of particular queries that are
associated with the views. This way, the various
annotations of resources are used as organizational
factors, depending on the user’s needs. For instance,
the user can create a dynamic view with the query
“Find all hairpins that produce mature with name
hsa-mir-147a”. This will organize into a dedicated
workspace all resources that are matched by the
evaluation of this query. Updating the dynamic view
will result in repopulating the view based on the
updated query result set.
3 RESOURCE SEARCH AND
EXPLORATION
Search and exploration in LinkZoo combines
keyword-based search with property-based faceted
browsing. Specifically, we have implemented an
“on-the-go” search mechanism that serves
suggestions based on the taxonomy of resource types
as well as the properties of resources. This type of
search is applied on LinkZoo resources that have
been imported to or shared with a user’s account.
Furthermore, we have implemented a “search–with-
query-candidates” mechanism that can be used to
query remote endpoints imported in LinkZoo. This
way, LinkZoo allows exploring, importing, and
annotating remote datasets. Therefore, by combining
the search functionalities, users can first find
relevant endpoints and then query them explicitly.
LinkZoo also provides exploration by faceted
browsing. In every folder shown to the user, the
system also shows all properties from triples with
the contained resources as subjects. Then, upon
selection of a property, the objects in the triples of
that property will be listed in the form of virtual
folders for further exploration. For instance, in a
folder that contains MP3 audio files, the property
mo:genre will be selectable for faceted browsing.
Then, the MP3 resources will be organized to
DATA2015-4thInternationalConferenceonDataManagementTechnologiesandApplications
234
Figure 2: Example of an RDF data graph.
Figure 3: One of the 5 possible Augmented Summary Graphs for the keywords MIMAT0000251, name and hasTarget for
the data graph shown in Figure 2.
virtual folders based on the distinct values of the
mo:genre
property. The keyword search and
property filtering methods can be combined and
applied in an exploratory manner.
3.1 Search with On-the-Go Suggestions
An interactive, two-step search process is
implemented for the exploration of the user’s
imported and shared data within LinkZoo. Natural
language query suggestions are given to the user
based on the input keywords and the resource types
they match. For instance, by typing “Research”, the
user will be prompted to select a suggestion from a
list that contains queries of the form “find URLs with
rdfs:seeAlso dbpedia:Research”, if a similar pattern
exists in the user’s data. Upon selection of a
suggested query, the relevant results will be shown
in a virtual folder. These can then be further
processed in bulk for annotation, moving, deleting
and other resource-specific operations. This kind of
search works incrementally, on the results
previously fetched. For example, after the user
selects “find URLs with rdfs:seeAlso
dbpedia:Research” and the relevant results are
fetched, further keyword exploration will suggest
queries based only on the current result set and not
on the whole set of user data. The above points can
be summarized in the following steps: (1) Each
keyword entered by the user is matched to objects of
the triples of the user’s resources. The distinct
predicate-object pairs that are matched are ordered
by their resource types. (2) The system feeds back
suggestions of the form “find {resource_type} with
{predicate} {object} “. For example, “find URLs
with rdfs:seeAlso dbpedia:Research”. (3) The user
selects a suggestion and the system builds a query
based on the resource type and the predicate object
pairs found in (1). (4) The system feeds back the
results of the query in (3) to the user. (5) The user
goes back to (1) and enters a new keyword in order
to refine the results.
RDFResourceSearchandExplorationwithLinkZoo
235
3.2 Search with Query Candidates
A key feature of LinkZoo is a search service that
assists the user to explore remote RDF data sources
and to retrieve RDF entities, which, in turn, can be
imported in LinkZoo. Given a set of keywords,
LinkZoo returns a set of candidate SPARQL queries
that try to capture user’s information need as
expressed by the keywords. Briefly, given a set of n
keywords, we perform the following steps: (1) For
each keyword
,
we retrieve all its matches
on
the RDF data graph. (2) We calculate all possible
combinations =
×
×⋯×
=
{ =
(
,⋯,
)
|
∈
∀ = 1, ⋯ , } of all the
matched elements
where =1,⋯,. (3) For
each combination ∈ that contains one matched
element
per keyword
, we create an augmented
summary graph
. (4) From each augmented
summary graph
, we generate the query pattern
graph
and finally (5) we translate each query
pattern graph
. into a SPARQL query. Next, we
elaborate on the details.
Let’s consider that we have an RDF dataset as
the one shown in Figure 2. The dataset is depicted as
an RDF data graph, where oval shape vertices
represent RDF entities, diamond shape vertices
represent RDF classes and rectangle shape vertices
represent literals. Similarly, dashed edges represent
entity-to-attribute properties, while solid ones
represent inter-entities properties. Let's us assume
that the user has provided the keywords
MIMAT0000251, name and hasTarget. The first step
is to match the keywords to elements in the RDF
data graph: (1) MIMAT0000251 matches to the
literal “MIMAT0000251” that is connected via the
property “accession” with an RDF entity of
“Mature” type, (2) name matches to the literal
“NAME” that is connected via the property
“change” met with RDF entity of type “Mature” and
to the entity-to-attribute property “name” met with
entities of type “Hairpin”, “Mature”, “Species” and
“Gene”, resulting in 5 possible matches, and (3)
hasTarget matches to the inter-entities property
“hasTarget” met with subject of type “Interaction”
and object of type “Transcript”. The second step is
to calculate all possible combinations of the matched
elements and for each combination c create the
augmented summary graph
. In this example,
there are 5 possible combinations. Let's examine one
combination, where the name keyword matches to
the literal “NAME”.
Augmented Summary Graph. The augmented
summary graph
is a combination of an aggregated
representation of the RDF data graph , enriched
with graph elements for each matched
element
, =1,⋯,. More specifically, all
entities from the RDF data graph that have the
same type of RDF class are represented by a vertex
labelled with the name of the RDF class. Similarly,
all inter-entities properties of the same type are
represented by a directed edge between the
aggregated vertex representation of the subjects and
the aggregated vertex representation of the objects.
The edge is also labelled with the property's name.
Note that entity-to-attribute properties as well as
literal values are omitted from the summary
representation. Overall, the augmented graph is
actually an abstraction of the RDF data graph .
The augmented summary graph
contains also
graph elements for each element
from the set of
matched elements . More specifically, if the
matched element
is a literal value, then the
graph is extended by a directed edge and a vertex.
The edge represents the entity-to-attribute property
that the matched element is met with in the RDF
data graph, while the vertex is the matched element
itself. Note that the edge is attached from the
aggregated vertex representation of the subject to the
newly inserted vertex. Similarly, if the matched
element
is an entity-to-attribute property, then
the graph is extended by a directed edge and a
vertex. The edge represents the entity-to-attribute
property, i.e. the matched element
, and it is
attached from the aggregated vertex representation
of the subject to the newly inserted vertex. The
difference from the previous case is that the latter
vertex represents the unknown literal of the
property. Note that if the same entity-to-attribute
property is met with multiple RDF entities of
different RDF types in the RDF data graph that
would lead to different sets . An example of the
Augmented Summary Graph for the combination
under investigation is shown in Figure 3.
Query Pattern Graph. In order to extract the
query pattern graph
from the augmented
summary graph
, we calculate the shortest paths
between every pair of matched elements and we
combine all of them into one connected subgraph.
Note that during the shortest path calculations we
ignore the directionality of the edges. Moreover,
since a matched element
, i.e. a source or sink of
the shortest path algorithm, can also be an edge, then
the distance between two matched elements counts
the number of both vertices and edges that needs to
traverse across the augmented summary graph
.
For the Augmented Summary Graph of Figure 3,
since we have three keywords, we need to calculate
three shortest paths. We then combine the shortest
DATA2015-4thInternationalConferenceonDataManagementTechnologiesandApplications
236
paths into a single connected component, resulting
to the query pattern graph shown in Figure 4. Note
that the extra node “Transcript” is attached to the
property “hasTarget” in order to form a complete
triple pattern, although it is not part of neither of the
shortest paths.
Candidate SPARQL Generation. The final
step of the mapping process is to translate the query
pattern graph
into a SPARQL query. Note that
the vertices of the query pattern graph
are either
known or unknown literal values and aggregated
representations. We need to connect the latter type
of vertices and the vertices of unknown literal values
with variables in order to form the SPARQL triple
patterns. Note that labels of the vertices can be used
as constants in the triple patterns, while the labels of
the edges as predicates. To produce conjunctive
SPARQL queries, given the above observations for
every vertex ∈
, we perform the following:
• if is a literal, do nothing
• if is an unknown literal, then connect the
vertex into a new variable ().
• If is a aggregated representation for entities of
RDF type class with label
(
)
= ,
then the vertex is connected into a new variable
(
)
and produce the following SPARQL
triple
(
)
rdf:type
(
)
.
Similarly, for every edge ∈
:
• If represents an inter-entities property between
a vertex and a vertex ,then we
produce the triple pattern
(
)
(
)
(
)
.
• If represents an entity-to-attribute property
between a vertex and a vertex
that is a literal, then we produce the triple
pattern
(
)
(
)
(
)
.
• If represents an entity-to-attribute property
between a vertex and a vertex
that is an unknown literal, then we produce the
triple pattern
(
)
(
)
(
)
.
In our example, the pattern graph of Figure 4 is
mapped to the following SPARQL query.
SELECT ?I ?M ?T WHERE
{?I a diana:Interaction.
?M a diana:Mature.
?T a diana:Transcript.
?I diana:hasMature ?M.
?I diana:hasTarget ?T.
?M diana:accession “MIMAT0000251”.
?M diana:change “NAME”.}
Figure 4: The Query Pattern Graph extracted from the
Augmented Summary Graph of Figure 3. In dashed style,
we depict the matched elements.
4 DEMONSTRATION
In this section we demonstrate the search capabilities
of our tool. We employ a use case taken from the
DIANA linked dataset. The dataset contains
aggregated information from well-known biology
databases, including ENSEMBL, miRBase and
KEGG pathway, of the microRNA world published
in RDF, available at the endpoint
http://leonardo.imis.athena-innovation.
gr:8891/diana/sparql.
Let us consider the following scenario: a user is
engaged in a bioinformatics research project which
concerns control mechanisms for cancer studies, and
more specifically it focuses on the regulatory
microRNA molecules. To this end, the user has
gathered resources and data from a variety of
sources, such as publications from PubMed, and data
from the Gene Expression Atlas, the Experimental
Factor Ontology and DIANA. Some of these
resources have been imported by the user himself,
while others have been shared to him by his
collaborators. Publications are modelled either with
the file or URL type and are annotated with metadata
provided by the user and collaborators as well as
external enrichment services. On the other hand, the
imported datasets are modelled as resources of the
type DataCollection, which allows the exploration
of a remote RDF dataset via our search-with-query-
candidates mechanism. Similarly to other resource
types, DataCollection resources are annotated with
descriptive metadata.
We consider that the user has either limited
knowledge of the RDF vocabulary used to describe
the datasets, or limited experience with SPARQL.
To overcome this problem, LinkZoo offers the
capability of keyword search for identifying and
exploring RDF datasets. The user can identify
RDFResourceSearchandExplorationwithLinkZoo
237
potential datasets that fit his criteria, based on
metadata annotations. In this case, he is looking for
datasets that contain microRNA data, and to that end
he uses the search-on-the-go utility to eventually
identify the DIANA dataset.
After identifying DIANA as the dataset to work
with, the user is interested in collecting information
about zebrafish miRNAs, in order to evaluate a
potential correlation with human cancer cell
metastasis. To achieve this, he types the words
“zebrafish hairpin” into the keyword search box and
as a result he gets two possible SPARQL queries.
Both generated queries will search for publications
that are annotated with the mesh term “zebrafish”
and are related with miRNAs of hairpin type. In the
first query, the “hairpin” keyword matches to the
RDF class
diana:Hairpin and imposes a direct
constraint that the property
diana:hasMirna of the
RDF class
diana:PaperMirnaConnection will
retrieve only Hairpin entities, while in the second
one the “hairpin” keyword matches to the literal
value of the property
diana:mirnaType of the
RDF class
diana:PaperMirnaConnection,
imposing an indirect constraint to the property
diana:hasMirna. Moreover, the first query will
also retrieve the RDF entities of the connected
Hairpins, while the second will not. The data he
retrieved from the keyword search request, could
provide useful information that would allow the user
to further explore the dataset. Also, the user can
retrieve results by selecting one of the generated
queries, and incorporate them as new resources in
his LinkZoo account, in order to annotate them and
share them with his collaborators.
4.1 Preliminary Evaluation
In order to evaluate the search with query
candidates, we perform an effectiveness study. We
have asked our biologists collaborators to provide
keyword queries along with a description in natural
language of the required information. We have
aggregated 5 queries for the DIANA dataset. An
example query is ““Alzheimer's disease” mature”
and the corresponding description is “Retrieve all
mature miRNAs that are related with Alzheimer’s
disease”. To evaluate the effectiveness of our
generated queries we order them in reverse order
given the number of triple patterns they contain and
we calculate the Reciprocal Rank metric defined as
RR = 1/r where r is the rank of the correct query.
Given our problem definition, a query is correct if it
matches the information needs as explained in the
provided natural language description. Figure 5
shows the Reciprocal Rank we have calculated for
the 5 queries for the DIANA dataset. In the 4 out of
5 queries, we got an RR of 1 meaning that we were
able to get the information required by the users.
Figure 5: Reciprocal Rank for the DIANA dataset.
5 RELATED WORK
Collaborative editing and annotating has been
explored and addressed thoroughly on the schema
level. Tools available for collaborative ontology
editing are presented in (Auer et al., 2006; Farquhar
et al., 1997; Tudorache et al., 2013). However, these
require expertise on the schema level. Regarding the
management of heterogeneous resources, Personal
Information Management (PIM) systems and tools
have been implemented, employing common
representation semantics as an abstraction layer
(Bernardi et al., 2011; Franz et al., 2007; Sauermann
et al., 2006). However, these address the
management of resources in non-collaborative
communities and are thus limited to individual
usage.
On the other hand, the keyword search problem
over structured data, tree structured (Cohen et al.,
2003; Kimelfeld and Sagiv, 2006) or graph
structured (He et al., 2007; Bhalotia et al., 2002), is a
problem that has widely been explored. Βasic steps
in those works involve 1) mapping the keyword
elements to data elements 2) searching for
substructures on the data that connect the keyword
elements and 3) return as output the substructures
given a scoring function. (Tran et al., 2009)
proposed a different solution to the keyword search
problem, where instead of computing for the
answers directly, it computes structured queries
allowing the user to choose the appropriate one.
LinkZoo’s approach on keyword search with query
candidates follows (Tran et al., 2009) approach to
generate SPARQL queries, but uses a different
exploratory method. In particular, we create multiple
augmented graphs one per keywords combination
DATA2015-4thInternationalConferenceonDataManagementTechnologiesandApplications
238
and use the notion of shortest paths to create a query
pattern graph.
6 CONCLUSIONS
In this paper, we have presented LinkZoo, an IT
infrastructure for collaborative management of
heterogeneous resources on the Web. LinkZoo
provides an environment for modelling and
publishing data such as files, websites, datasets and
people as RDF, and allows for their coexistence in
shared contexts. Furthermore, we have presented the
various types of search capabilities implemented in
the platform. These span from trivial text searching
within a user’s data to more elaborate data-guided
exploration and searching over imported RDF data
collections. Finally, we have demonstrated the
usability of the search functions through a use case
drawn from the life sciences domain.
Currently, keyword search expects exact
matches of terms. In the future, we will extend this
functionality to automatically suggest terms from the
RDF data graph. Another direction will be to extend
the matching procedure by enabling also ontology
matching (Euzenat and Shvaiko, 2013).
Furthermore, to assist user understanding of the
produced candidate SPARQL queries, we intend to
also show natural language descriptions of the
generated candidates. Finally, we also plan to
perform an extensive evaluation of our search
services, in term of completeness of the results, time
and memory requirements for indices creation,
performance.
ACKNOWLEDGEMENTS
This study has been partially supported by
LODGOV project, Research Programme ARISTEIA
(EXCELLENCE), General Secretariat for Research
and Technology, Ministry of Education, Greece and
the European Regional Development Fund, and the
Operational Program "Education and Lifelong
Learning" of the National Strategic Reference
Framework (NSRF) - Research Funding Program:
Thales. Investing in knowledge society through the
European Social Fund.
REFERENCES
Alchemy API, 2015 Available from: <http://www.
alchemyapi.com/>. [2015].
Auer, S., et al. 2006, ‘OntoWiki–a tool for social,
semantic collaboration’, Proceedings of ISWC, ed.
Springer Berlin Heidelberg, pp. 736-749.
Bernardi, A., et al. 2011, ‘The NEPOMUK semantic
desktop’, Proceedings of CSKM, Springer Berlin
Heidelberg, pp. 255-273.
Bhalotia, G., et al. 2002, ‘Keyword searching and
browsing in databases using BANKS’, Proceedings
of ICDE, pp. 431-440.
Bizer, C., et al. 2009, ‘Linked data-the story so far’, Int. J.
Semantic Web Inf. Syst., vol 5, no 3, pp.1-22.
Bizer, C., & Schultz, A. 2008, ‘Benchmarking the
performance of storage systems that expose SPARQL
endpoints’, Proceedings of WWW.
Cohen, S., et al. 2003, ‘XSEarch: A semantic search
engine for XML’, Proceedings of VLDB, pp. 45-56.
Euzenat, J., & Shvaiko, P. 2013, Ontology matching (2
nd
edition), ed. Heidelberg: Springer-Verlag.
Farquhar, A., Fikes, R., & Rice, J. 1997, ‘The ontolingua
server: A tool for collaborative ontology construction’,
International journal of human-computer studies, vol.
46, no. 6, pp. 707-727.
Faye, D. C., Cure, O., & Blin, G. 2012, ‘A survey of RDF
storage Approaches’, ARIMA Journal, vol.15, pp.11-35.
Franz, T., Staab, S., & Arndt, R. 2007, ‘The X-COSIM
integration framework for a seamless semantic
desktop’, Proceedings of K-CAP, pp. 143-150.
He, H., et al. 2007, ‘BLINKS: ranked keyword searches
on graphs’, Proceedings of SIGMOD, pp. 305-316.
Kimelfeld, B., & Sagiv, Y. 2006, ‘Finding and
approximating top-k answers in keyword proximity
search’, Proceedings of SIGMOD-SIGACT-SIGART
pp. 173-182.
Meimaris, M., Alexiou, G., & Papastefanatos, G. 2014,
‘LinkZoo: A linked data platform for collaborative
management of heterogeneous resources’,
Proceedings of ESWC, pp. 407-412.
Prud’Hommeaux, E., & Seaborne, A. 2008, ‘SPARQL
query language for RDF’, W3C recommendation.
Sauermann, L.,et al. 2006, ‘Semantic desktop 2.0:The
gnowsis experience’, Proceedings ISWC, pp.887-900.
Tran, T., et al. 2009, ‘Top-k exploration of query
candidates for efficient keyword search on graph-
shaped (rdf) data’, Proceedings of ICDE, pp. 405-416.
Tudorache, T., et al. 2013, ‘WebProtégé: A collaborative
ontology editor and knowledge acquisition tool for the
web’, Semantic web, vol. 4, no. 1, pp. 89-99.
RDFResourceSearchandExplorationwithLinkZoo
239
