=Paper=
{{Paper
|id=None
|storemode=property
|title=Integrating Ecological Data Using Linked Data Principles
|pdfUrl=https://ceur-ws.org/Vol-938/ontobras-most2012_paper13.pdf
|volume=Vol-938
|dblpUrl=https://dblp.org/rec/conf/ontobras/MouraPPPMV12
}}
==Integrating Ecological Data Using Linked Data Principles==
https://ceur-ws.org/Vol-938/ontobras-most2012_paper13.pdf
Integrating Ecological Data Using Linked Data Principles
Ana Maria de C. Moura1, Fabio Porto1, Maira Poltosi1 , Daniele C. Palazzi1, Régis
P. Magalhães2, Vania Vidal2
1
Extreme Data Lab (DEXL Lab)
National Laboratory of Scientific Computing (LNCC)
Petrópolis – RJ – Brazil
2
Department of Computing
Federal University of Ceará (UFC)
{anamoura, fporto, maira, dpalazzi}@lncc.br, {regispires,
vvidal}@lia.ufc.br
Abstract.This paper presents a framework to manage, treat and integrate ecological
data in the context of the PELD project, currently in development in Brazil. These data,
which are produced and collected from different resources, are stored in distinct
relational databases and transformed later into RDF triples, using a traditional
relational-RDF mapping. Taxonomical, spatial and trophic relations are explored by
means of ontological properties, which make it possible to discover interesting
information about existing marine species of different bays in the country, illustrated by
SPARQL queries. Additionally, the endpoint thus generated allows data to be accessed
on the Web of data, as linked data.
1. Introduction
Extensive information on policies, action programs, and environmental challenges in
areas such as sustainable development, climate change, environmental law, and
biodiversity, has become a great concern throughout the world. Different governmental
agencies1 and commissions2,3 have been created for the purpose of defining strategies to
preserve natural environment. Among these many policies, there is a strong concern on
developing systems to organize and catalogue information about the existing natural
reserves, such as minerals and biological ones, which involve fauna, flora and hydro
resources, enabling a more accurate control of this information.
In Brazil, a great effort is being deployed in this direction through an important
national project named PELD/Brazil4 (Brazilian Long-Term Ecological Research
Program). One of its main goals is to leverage ecological knowledge, so that important
data can be provided to help, reinforce government decisions, and support research
related to the management of natural resources, as well as to share this information
among different sectors of society. PELD project currently counts on 29 collect sites,
which are distributed along different Brazilian biomes, for the purpose of consolidating
the existing knowledge about their composition and learning about ecosystems
functioning. Having an integrated view of these ecological data sources and making
1
http://www.environment-agency.gov.uk/
2
http://ec.europa.eu/environment/index_en.htm
3
http://www.princetontwp.org/environmain.html
4
http://ppbio.inpa.gov.br/Port/projetosassociados/peld/
156
them available on the Web of data as a data set [Heath, Bizer 2011], would permit other
ecologist researchers throughout the world to access, as well as to reference it to other
data sets, dealing with similar subjects.
A PELD site can be considered as an integration of many sub-projects
concerning distinct ecological issues. Since most of these PELD sites throughout the
country are not still consolidated, or are in an initial development phase, in this paper
we focus on the Guanabara PELD5. This PELD site aims at extending knowledge about
the Guanabara Bay ecosystem and providing support for managing, structuring and
publishing ecological data, as well as to be a source of answers to the anthropic and
climatic impacts on the bay ecosystem. Currently a database project is being developed
to manage, organize and access information about Guanabara Bay ecological data.
However, since Guanabara PELD is developed by a large group of biologists,
responsible by distinct domains (hydrology, planktons, fishes, ecology, etc.), data are
produced independently, in different formats, and according to specific methodologies.
Integrating and publishing all the data produced by these groups is crucial not only to
provide a homogeneous view of this data, but also to make it available for other groups
working in other PELDs throughout the country. This situation offers an interesting
panorama to evaluate how efficient queries and reasoning will be in the face of a query
federation pattern, where data are integrated according to the Linked Data (LD)
strategy.
The main contribution of this work in comparison to other existing ecological
information management systems (Ecoflora6 [Cavalcanti 2005]) is: to integrate different
ecological resources and to make them available on the Web of data, using LD
principles; to provide reasoning capacity, i.e., to infer new information from the stored
data. By providing an ontological representation of the data model, new relationships
and instances may be inferred, taking into account transitive properties and hierarchies
over the model concepts, allowing researchers to discover interesting data, such as, for
example, information about specie’s predators in different levels of a hierarchy.
In this paper we extend the integration framework [Vidal et al. 2011] and use
some techniques to create the application ontologies. Query results are extracted from
PELD data sources, integrated by QEF7 framework [Porto et al 2007], and then
visualized by the user as linked data.
The remainder of this paper is structured as follows. Section 2 presents related
work. Section 3 presents the framework architecture designed to integrate PELD
resources. Section 4 describes some PELD application scenarios that will be used as
study case for integration. Section 5 describes the scenario ontologies generated at each
level of the proposed architecture, as well as the mappings rules between the domain
and application ontologies. Section 6 shows how to answer user´s queries in this
architecture as linked data, by presenting a query example over different PELD
resources, executed in SPARQL. Finally, Section 7 concludes the paper with
suggestions for future work.
5
http://www.lncc.br/peldguanabara/index.php
6
http://www.ecoflora.co.uk/
7
http://146.134.234.248/QEF/index.html
157
2. Related Work
Several works such as Ecoflora8, NRCS8, AEZ9, [Cavalcanti 2005], [Campos et al.
2009], [Manzi 2009] have been proposed to manage and share ecological data.
However, they do not perform data integration on LD using multiple data sources. They
also do not address inference provided by the use of ontologies or semantic web
approaches. This paper intends to fill this gap, by proposing an integration approach
based on LD, which enables ecological data to be analyzed, inferred and queried from
different PELD sources. Below we present related works that address data integration
over LD.
There are two possible approaches for data integration: materialized and virtual.
The first approach collects, stores and accesses data in a central database. The main
disadvantage of this approach is the replication of data, which in addition requires
additional storage space and does not ensure the use of updated data in relation to the
original datasources. LDIF [Schultz et al. 2011] is a framework that provides data
integration through the use of the materialized approach. On the other hand, the virtual
approach enables the execution of federated queries over a fixed set of datasources. Our
work uses both the materialized and the virtual data integration approach. Jena ARQ10
SPARQL, DARQ [Quilitz and Leser, 2008], SemWIQ [Langegger, 2010] and FedX
[Schwarte 2011] are examples of systems that provide transparent access to RDF data
sources, whose data can be retrieved using SPARQL. While some of them, such as
SemWIQ, allows RDF schema or OWL ontologies to be used to describe the
datasources, FedX transforms the original query into a federated query over the source
ontologies. However, none of these tools can execute queries over a domain ontology
with mappings for specific application ontologies.
The integration of scientific data in the context of linked data using the virtual
approach has been discussed in [Gray et al. 2008]. In that paper the authors discuss the
integration of astronomic databases using RDF as a common schema language and
SPARQL as a query language. The authors adopt a peer-to-peer integration strategy,
avoiding a global view agreement. In the proposed view, each database is exposed in
RDF and alignment mappings define associations between databases.
The integration tools presented in this section require the manual definition of
the datasources used in each query. It is also necessary to rewrite queries when a
datasource schema changes. However, the generation of federated query plans from
queries over a Domain Ontology can accomplish the semantic integration in virtual and
automatic way. Queries over the domain ontology are also simpler and more stable than
if they were made directly over the application ontologies.
3. Integration Architecture
The need to produce data in PELD projects in a homogenous format is a fundamental
requirement when considering the generation of an integrated view of Brazilian
ecosystems. In this context, RDF (Resource Description Framework) [Manolla, Miller
2004] has been used as a powerful strategy to interoperate, reason and publish data,
besides enabling these data to connect with
8
http://plants.usda.gov/java/
9
http://www.fao.org/nr/land/databasesinformation-systems/aez-agro-ecological-zoning-system/en/
10
http://jena.apache.org/documentation/query/
158
other resources of similar domains. Additionally, it enables the exploration and
association among data, making use of SPARQL [Prud´hommeaux, Seaborne 2008].
Nevertheless, although the great benefits of RDF, there is a great concern when
using it to deal with large volumes of data, since it may degrade performance [Gray et
al. 2009]. This is why a current adopted strategy is to store data in relational databases.
Moreover, publishing data according to the Linked Data best practices [Heath, Bizer
2011] solves part of the integration problem, which is to make data available in a
common format. Ontologies come as a rescue ground from which integration becomes
possible. They provide a common vocabulary to be shared among the different data
sources. Thus, one needs to combine the publication of source data according to the LD
best practices using RDF with a common shared vocabulary expressed as a domain
ontology.
Figure 1 presents a three-level architecture used to integrate relational schemas
as LD. It is based on mappings according to a mediated approach, and it has been
extended from [Vidal et al 2011] to integrate the PELD databases as described below.
Figure 1. Three-level architecture for Linked Data Integration
The RDF Domain Integration View (1) is the Domain Ontology (DO) that
represents the mediated schema. Designed by an expert user, it provides a conceptual
representation of a specific domain, which comprises a global shared vocabulary and
constraints. Each PELD relational database (5) is transformed into RDF by a specific
wrapper (4) (see section 4.4) and becomes a source ontology (3), which is then rewritten
as a PELD application ontology (APO) (2). It is worth observing that each APO
describes a source ontology according to the principles of LD, which is a subset of the
DO ontology. Application ontologies help breaking the query answering problem in two
steps: (i) a query is submitted to the mediated schema, i.e., to the domain integration
view, and by using mediated mappings, the query over the integration view is rewritten
in terms of the application ontologies. As an example, consider queries over the Sample
concept (Figure 2), which are rewritten as unions of AO; then (ii) based on the rewritten
query an execution plan is generated, in which references between APOs become joins,
and each sub-query, completely covered by an AO is rewritten using local mappings,
159
and then submitted to the corresponding PELD local databases to retrieve information
and deliver an integrated query answer to the user as LD. This step by step procedure is
better described in section 6.
4. Application Scenarios
This section describes the Guanabara PELD scenarios that will be used for integration,
based on the architecture depicted in Figure 1.
Guanabara PELD aims at getting biotic data from samples extracted from the
bay water and from fishing resources. The living organisms are hierarchically classified
in a taxonomy. The first level corresponds to the Kingdom, which is decomposed into
Phylums and successively into classes, orders, families, genders and species. Each
level has respectively its own subdivisions. Any level within this classification is called
a taxon. There exist differences in the levels concerning each organism. Some of them
have been reclassified, and in this case, both classifications are kept, and a synonymous
relation is established between them.
In the context of ecological data analysis, some important features deserve some
attention. Geographical region information identifies a target ecosystem and is used for
selecting and classifying events according to their location. On the other hand, trophic
relations are fundamental for the ecosystem study. Finally, the taxonomy enables a
hierarchical analysis of the species. The analysis of these aspects may be explored by
the use of inference in an integrated way.
The main characteristics of each scenario are described next.
• Plankton: in the plankton scenario, a sample data takes into account temporal (data
and time) and spatial (latitude, longitude and profundity) information, as well as
methods used for sample collect and conservation, atmospheric, and maritime
conditions during each collect. For each analysis performed, data, sample and the
applied method are registered. Biomass measurements of organisms found in the
samples can be done at specie level or at the taxonomy highest level;
• Community Fish: besides temporal and spatial information, this application scenario
stores the fishing method used to catch fishes, taking into account two different
depths (initial and final). It is worth observing that collected fishes are divided into
three samples, from which the total weight and number of individuals are analyzed
for each taxon found in the collect process;
• Catfish Genidens: differently from the previous scenarios, this application scenario
analyzes each specific specie individually, considering not only spatial and temporal
references, but also the fishing method employed in the collect process, the specie
weight, length and gender.
5. Domain and Application Ontologies
Based on the application scenarios described above, this section describes the
ontologies generated at each level of the framework architecture presented in Figure 1.
Domain Ontology(DO)
Since in this paper the main purpose is not ontology design, we assume the domain
ontology is provided by the user. Figure 2 presents the conceptual representation of the
PELD domain ontology, referenced in our architecture as RDF Domain Integration
160
View. The namespace prefix “d” is used to refer to the vocabulary of this domain
ontology. Since most of the class properties are self-described, we just give a few
examples of the class properties. Thus, d:collect_method is defined as a datatype
property with domain d:Sample and range string; d:has_predator is an object type
property with domain d:Trophic_Chain and range d:Taxon; and d:has_pl_analysis is
also defined as an object property, with domain d:Plankton_sample and range
d:Pl_analysis.
Application Ontology(AO)
As mentioned in section 1, PELD sites are composed of different PELD subprojects.
Each such PELD subproject takes part in the PELD data integration, by providing their
local data published in RDF, which is rewritten as an AO, using a subset vocabulary of
the DO. As in a federated database, an application ontology may be seen as an external
ontology that takes part in the integrated schema, i.e., the domain ontology. Figure 3
presents a conceptual representation of the PELD AOs associated with the application
scenarios described above comprising five ontologies: Plankton, Catfish Genidens,
Community Fishes, Region and Taxon, each one having the following namespace
prefixes: “apl:”, “acf:”,“aco:”, “r”, and “tx” respectively. As mentioned before, the
vocabulary of an application ontology consists of classes and properties that are subset
of the domain ontology. Thus, access to the local data is done through direct mappings
and the integration work becomes facilitated.
Based on the work proposed in [Vidal et al 2011], Figure 4 presents the list of
the rules defined for the mapping between the APO and the DO. Due to space
restriction we present only the mapping rules of Plankton ontology and we refer the
reader to the above reference for more details on the definition of these rules, which is
not in the scope of this paper.
It is worth mentioning that since the ontologies Region and Taxon represent data
that are not frequently changed, they are previously materialized and stored locally as
RDF triples in a repository, also as AOs. Thus, they are accessed whenever required and
joined together with the other APOs that are virtually retrieved, as described in section
6.2.
6. Querying over the Framework Architecture
The main purpose of the proposed integration framework architecture is to answer
user´s queries in terms of a domain ontology. Through the unified view exposed by the
DO, researchers can access PELD subproject data transparently independently of local
particularities. In order to deliver data, the data integration framework must be
supported by a data integration engine that processes user´s query requests and returns
results dealing with necessary data translations and access to source data11. In the
context of this paper, ontologies in all architecture levels are homogeneously expressed
in RDF. Thus, user requests may be submitted to the data integration system using
SPARQL. The query expression is transformed into sub-queries over the application
ontologies exposed as RDF triples by the D2RQ [Bizer et al. 2006] engine from the
source databases.
The QEF system developed at DEXL laboratory has been used as the data
integration engine. QEF is an extensible query engine that supports user-defined
11
In the current version QEF does not rewrite queries yet. This is considered as a future work.
161
algebras and data structures. In order to support PELD data integration, a new version
named QEF-LD [Magalhães 2012] has extended QEF. This new version includes linked
data algebraic operators, and wrappers that submit AO sub-queries to a D2R endpoint.
The latter exports local databases as virtual AOs.
Figure 2. PELD domain ontology
In scenarios where a domain ontology query is translated into sub-queries over
more than one application ontology, results are combined by the Union operator and
returned to the user in a single result set.
Considering the strategy developed in [Vidal et al 2011], the following
algorithm is performed:
- The user submits a SPARQL query to the data integration system expressed in terms
of a domain ontology. Then, according to the mediated mappings, an integrated
query execution plan is generated according to the following steps:
a. References to the concepts Region and Taxonomy in the query, which are shared
by the AOs, are mapped to BindJoins [Magalhães 2012] between the source AO
and the shared databases (i.e. Region or Taxonomy).
b. Each sub-query is submitted to a data source. D2R endpoints translate the
submitted queries to the corresponding local database queries. The Region and
Taxonomy AOs are materialized as RDF sources and joined with AO ontology
concepts through SPARQL queries.
c. Once the results are obtained, QEF applies the joins and unions handling in the
final result. A query over the Sample concept is rewritten as Unions of
subqueries over each APOs (see figure 4 (a)), according to the mappings
presented in Figure 4(b), respectively. This step is not currently supported by
QEF-LD [Magalhães 2012].
162
Figure 3. PELD application ontologies
d:Sample(p) ⇐ apl: Plankton_Sample(pls) U acf: Catfish_Sample(c) U
aco: Comm_Fish_Sample(co)
Figure 4(a). Sample DO expressed as the union of the different sample species of
APO ontologies
1. d:Pl_analysis(pl) ⇐ apl:Pl_analysis(pl)
2. d: Plankton_Sample(pls) ⇐ apl: Plankton_Sample(pls)
3. d:id_sample(p,id) ⇐ apl: id_sample(pls,id), apl: Plankton_Sample(pls)
4. d:col_date(p,dt) ⇐ apl: collect_date(pls,dt), apl:Plankton_Sample(pls)
5. d:collect_method (p,cm) ⇐ apl: collect_method(pls,cm), apl: Plankton_Sample(pls)
6. d:depth (pls,d) ⇐ apl:depth(pls,d), apl: Plankton_Sample(pls)
7. d:tide_condition (pls,tc) ⇐ apl: tide_condition(pls,tc), apl: Plankton_Sample(pls)
8. d:collect_in (p,l) ⇐ apl: collect_in(pls,l), apl:Plankton_Sample(pls), Region(l)
Figure
9. d:analysis_meth 4(b).⇐Mapping
(pl,am) rules from the
apl: analysis_meth(pl,am), Plankton APO to DO
apl:Pl_analysis(pl)
10. d:weight (pl,w) ⇐ apl: weight(pl,w), apl:Pl_analysis(pl)
11. d:has_taxon (pl,tx) ⇐ apl:has_taxon(pl,tx) , apl:Plankton_analysis(pl), Taxon(tx)
12. d:has_pl_analysis (p,pl) ⇐ apl:has_pl_analysis(pls,pl)
13. d:type (p,´plankton`) ⇐ apl: Plankton_Sample(p)
Figure 4(b). Mapping rules from the Plankton APO to DO
6.1 Submitting a Query
According to the strategy presented above, the following query has been submitted to
the proposed framework: “Get the species found at Paquetá Island in 2004, their
synonyms and predators”. In the following paragraphs the transformation process for
answering this query is described, step by step.
i) The main query (Q) is expressed in terms of the domain ontology, which
comprises the union of the 3 species: Planktons, Catfish and Comm. Fish.
163
Select distinct ?name ?name_syn Union {
?name_pred ?p d:is_a ?s .
Where { ?p d:has_cf_analysis ?pl .
{ ?pl d:id_taxon ?tx .
?s d:collected_in ?r . ?tx d:popular_name ?name .
?s d:collect_date ?dt. }
?r d:name ?reg. Optional {
?p d:is_a ?s . ?tx d:has_predator ?pred .
?p d:has_pl_analysis ?pl . ?pred d:has_taxon ?idpred .
?pl d:id_taxon ?tx. ?idpred d:popular_name ?name_pred .
?tx d:popular_name ?name }
} Optional {
Union { ?syn d:is_synonimous-of ?tx;
?p d:is_a ?s . d:popular_name ?name_syn .
?cf d:id_taxon ?tx . }
?tx d:popular_name ?name . Filter (?reg = "Paqueta" && ?dt = 2004 )
} }
order by ?name
ii) Query Q is rewritten as the union of three subqueries Q1, Q2 and Q3, which aim
at extracting data from Plankton, Catfish Genidens, and Comm. Fish application
ontologies, region and taxon, respectively (Figure 5).
6.2 Executing a Query in QEF
As mentioned before, part of the application ontologies are stored in RDF tuples as
materialized views. Such characteristic requires an execution plan for each query Qi
(Figure 6(a)), in order to guide QEF into the correct execution of the algebra operators
sequence over the local data sources.
In order to exemplify this step, consider query Q´1 the Q1 version about
Planktons that will be submitted to QEF. Similarly to Q2 and Q3, these queries use both
virtual and materialized information. In other to describe the step by step execution
procedure performed by QEF, Figures 6 (a) and (b) present respectively a Q´i query
execution plan for each Qi, and each corresponding SPARQL query. Figures 7, 8, and 9
present, respectively, the results of Q´1, Q´2 and Q´3.
Final results (Figure 10) are obtained from Q´1 ∪ Q´2 ∪ Q´3, having duplicated
values discarded.
7. Conclusion
This paper reports on the application of the aforementioned data integration framework
to the ecological domain and the extension of QEF, a data integration system, to answer
queries on heterogeneous ecological databases using this framework. A complete data
integration scenario is discussed based on the challenges involved in publishing
ecological data produced by the PELD Guanabara project, in Brazil.
Based on the data integration framework, a set of PELD subproject databases
stored in relational databases are transformed into RDF as endpoints via D2RQ, which
enable an integrated view over the data resources via SPARQL queries. The results
indicate that the proposed data integration framework is promising and that shall be
adopted as a standard for more complex ecological database integration scenarios.
164
Q1 Q2 Q3
Select distinct ?name Select distinct ?name Select distinct ?name ?name_syn
?name_syn ?name_pred ?name_syn ?name_pred ?name_pred
Where { Where { Where {
?s apl:collected_in ?r. ?s acf:collected_in?r. ?s aco:collected_in ?r.
?s apl:col_date ?dt. ?s acf:col_date ?dt. ?s aco:col_date ?dt.
?r r:name?reg. ?r r:name ?reg. ?r aco:name?reg.
?s apl:has_pl_analysis ?a. ?s tx:has_taxon ?tx. ?s aco:has_cf_analysis ?a.
?a tx:has_taxon ?tx. ?tx tx:popular_name ?name. ?tx tx:popular_name ?name.
?tx tx:popular_name Optional { ?a tx:has_taxon ?tx.
?name. ?tx tx:has_predator ?pred. Optional {
Optional { ?pred tx:has_taxon ?idpred . ?tx tx:has_predator ?pred.
?tx tx:has_predator ?pred. ?idpred tx:popular_name ?pred tx:has_taxon ?idpred.
?pred tx:has_taxon ?idpred. ?name_pred. ?idpred tx:popular_name
?idpred tx:popular_name } ?name_pred.
?name_pred. Optional { }
} ?syn tx:is_synonimous_of Optional {
Optional { ?tx; ?syn tx:is_synonimous_of ?tx;
?syn tx:is_synonimous_of tx:popular_name ?name_syn tx:popular_name ?name_syn
?tx; } }
tx:popular_name Filter (?reg = "Paqueta" && Filter (?reg = "Paqueta" && ?dt
?name_syn}. ?dt = 2004 ). = 2004 ).
Filter (?reg = "Paqueta" && } }
?dt = 2004). order by ?name order by ?name
} order by ?name
Figure 5. Qi SPARQL query
Figure 6(a). Q´i execution plan
Q´1 (Planktons) Q´2 (Catfish) Q´3 (Comm.fish)
Qplankton: Select ?id_taxon, Qcatfish: Select ?id_taxon, Qcommfish: Select ?id_taxon,
?id_region ?id_region ?id_region
Where { Where { Where {
?s apl:collected_date ?dt. ?s acf:collected_date ?dt. ?s aco:collected_date ?dt.
?s apl:collected_in ?id_region. ?s acf:collected_in ?id_region. ?s aco:collected_in ?id_region.
?s apl:has_pl_analysis ?id_an. ?s tx:has_taxon ?id_taxon. ?s aco:has_cf_analysis ?id_an.
?id_an tx:has_taxon ?id_taxon. Filter (?dt =2004 ). ?id_an tx:has_taxon ?id_taxon.
Filter (?dt =2004 ). } Filter (?dt =2004 ).
} }
--------------------------------------- --------------------------------------- ----------------------------------------
Qregion: Select ?id_region Qregion: Select ?id_region Qregion: Select ?id_region
Where { Where { Where {
165
?id_region r:name ?n. ?id_region r:name ?n ?id_region r:name ?n
?r r:id_region ?id_region. ?r r:id_region ?id_region. ?r r:id_region ?id_region.
Filter (?n, “Paquetá”).} Filter (?n, “Paquetá”).} Filter (?n, “Paquetá”).}
---------------------------------------- --------------------------------------- ----------------------------------------
QTaxon: Select distinct ?name QTaxon: Select distinct ?name QTaxon: Select distinct ?name
?name_syn ?name_pred ?name_syn ?name_pred ?name_syn ?name_pred
Where { Where { Where {
?x tx:id_taxon ?id_taxon. ?x tx:id_taxon ?id_taxon. ?x tx:id_taxon ?id_taxon.
?id_taxon tx:scientific_name ?id_taxon tx:scientific_name ?id_taxon tx:scientific_name
?name. ?name. ?name.
?id_taxon tx:has_predator ?pred. ?id_taxon tx:has_predator ?pred. ?id_taxon tx:has_predator ?pred.
?pred tx:has_taxon ?tx_pred. ?pred tx:has_taxon ?tx_pred. ?pred tx:has_taxon ?tx_pred.
?tx_pred tx:scientific_name ?tx_pred tx:scientific_name ?tx_pred tx:scientific_name
?name_pred. ?name_pred. ?name_pred.
?id_taxon tx:is_synonimous_of ?id_taxon tx:is_synonimous_of ?id_taxon tx:is_synonimous_of
?syn_tax. ?syn_tax. ?syn_tax.
?syn_tx tx:scientific_name ?syn_tx tx:scientific_name ?syn_tx tx:scientific_name
?name_syn. ?name_syn. ?name_syn.
} } }
Figure 6(b). Q´i SPARQL query
Figure 8. Results of Q´2
Figure 7. Results of Q´1
Figure 10. Final Results
Figure 9. Results of Q´3
C
ACKNOWLEDGEMENTS
This work has been partially supported by CNPq through its Institutional Capacity
Program (Proc. 382.489/09-8) and Productivity Research fellowship (Proc.
309502/2009-8).
166
References
Bizer C., Health T., Berners-Lee T. D2R Server – Publishing relational databases on the Web as
SPARQL endpoints. Proc. of the 15th International World Wide Web Conference,
Edinburgh, Scotland, 2006.
Campos, S.R., Martinhago A.Z., Massahud R.T., França A.M., Prieto L. E., Mendes J.D.C.
Database modeling of the economic ecological zoning of Minas Gerais using UML-
GeoFrame (in Portuguese). Proc. of the XIV Brazilian Symposium of Remote Sensoring,
Natal, Brasil, 25-30 April, 2009, INPE, p. 4943-4949.
Cavalcanti, M. J. Database on Amazon biodiversity: experience on the Biotupé project. Biotupé:
Physical environment, biological diversity and sociocultural of Low Negro River, central
Amazon (in Portuguese). Santos-Silva, Aprile, Scudeller, Editora INPA, Manaus, 2005.
Gray A. J. G., Gray N., Ounis I. Can RDB2RDF tools feasibily expose large science archives
for data integration? The Semantic Web: Research and Applications – LNCS, Vol.
5554/2009, 491-505, 2009. DOI: 10.1007/978-3-642-02121-3_37.
Heath, T., Bizer C. Linked Data: evolving the Web into a global data space (1st edition).
Synthesis lectures on the semantic Web: theory and technology, 1:1, 1-136. Morgan &
Claypool ed., 2011.
Langegger, A., Wöß, W., Blöchl, M. 2008. A Semantic Web Middleware for Virtual Data
Integration on the Web. In: Proceedings of the 5th European Semantic Web Conference
(ESWC). Volume 5021 of Lecture Notes in Computer Science. Springer Verlag, pp. 493–
507.
Magalhães, R. P. Um Ambiente para Processamento de Consultas Federadas em Linked Data
Mashups. M.S. thesis, Universidade Federal do Ceará, 2012.
Manola, F. and Miller, E. RDF primer. W3C Recommendation, February, 2004. Available at:
http://www.w3.org/TR/rdf-primer.
Manzi, A. Data management of Brazilian long-term ecological research projects (in
Portuguese), research project, Edital MCT/CNPq Nº 59/2009 – PELD support proposals,
2009.
Porto F., Tajmouati O., Silva V. F. V., Schulze B., Ayres F. V. M. QEF - supporting complex
query applications, 7th IEEE International Symposium on Cluster Computing and the Grid —
CCGrid 2007 , Rio de Janeiro, Brazil, pp. 846-851.
Prud´hommeaux, E. and Seaborne, A. 2008. Sparql Query Language for RDF. W3C
Recommendation. Available at: http://www.w3.org/TR/rdf-sparql-query/.
Prud’hommeaux, E. And Buil-Aranda, C. SPARQL 1.1 Federated Query.
http://www.w3.org/TR/sparql11-federated-query/, 2011.
Quilitz, B. and Leser, U. 2008. Querying Distributed RDF Data Sources with SPARQL. In:
Proceedings of the 5th European Semantic Web Conference (ESWC). Volume 5021 of
Lecture Notes in Computer Science, Springer Verlag, pp. 524–538 (2008).
Schultz, A., Matteini, A., Isele, R., Bizer, C., and Becker, C. LDIF : Linked Data Integration
Framework. In Proceedings of the 11th Interational Semantic Web Conference ISWC2011.
pp. 1–4, 2011.
Schwarte, A., Haase, P., Hose, K., Schenkel, R., and Schmidt, M. Fedx: optimization
techniques for federated query processing on linked data. In Proceedings of the 10th
international conference on The semantic web - Volume Part I. ISWC’11. Springer-Verlag,
Berlin, Heidelberg, pp. 601–616, 2011.
Vidal, V.M.P, Macêdo, J.A.F., Pinheiro, J. C., Casanova, M. A., Porto F. Query processing in
a mediator based framework for linked data integration. IJBDCN 7(2): 29-47, 2011.
167