Extending SMW+ with a Linked Data Integration Framework

Abstract

In this paper, we present a project which extends a SMW+ semantic wiki with a Linked Data Integration Framework that performs Web data access, vocabulary mapping, identity resolution, and quality evaluation of Linked Data. As a result, a large collection of neurogenomicsrelevant data from the Web can be exibly transformed into a uni ed ontology, allowing uni ed querying, navigation, and visualization; as well as support for wiki-style collaboration, crowdsourcing, and commentary on chosen data sets.

Keywords

• Linked Data
• Data Integration
• Semantic MediaWiki

Introduction

of the project will bring ABA, Uniprot7, KEGG Pathway8, PharmGKB9 and Linking Open Drug Data [ 5 ] data sets together in order to solve the challenge of nding drugs that target elements within a disease pathway, but are not yet used to treat the disease. The genes associated with the found drugs may then be compared for commonalities through an integrated analysis of the related ABA structure expression data. 2

 

In tegrating Linked Data from the Web In this c hapter, we discuss the architecture of the SMW-LDE deployment, which is depic ted in gure 1.

Application  Layer  

SMW+  Semantic  Enterprise  Wiki    

Data  Access,    

Integration  and    

Storage  Layer  

Web  of  Linked  Data  

Publication  Layer  

    LDimporter,   LDspider  

Web  Data   Access  Module  

HTTP  

SPARQL  

SMW+  Linked  Data  Integration  Framework   R2R   Vocabulary   Mapping   Module  

Silk  Server   Identity   Resolution   Module  

TPEE   Quality   Evaluation   Module  

Integrated  

Web  Data  

HTTP  

HTTP  

HTTP  

RDF/   XML   ABA   RDF/   XML  

Uniprot  

D2R  Server   D2R  Server   KEGG  

Pathway  

PharmGKB  

All base data sets for the project, including Brain Atlas data, are published on the Web according to the Linked Data principles, thereby becoming part of a giant global graph { the Web of Linked Data. This logical graph is depicted in the Web of Linked Data layer in gure 1. Non-RDF data sources such as KEGG Pathway and PharmGKB are published using D2R Server10. The SMW+ Linked Data Integration Framework takes the role of a data access, integration and storage layer that makes Web data available to the application (in this case, 7 http://www.uniprot.org/ 8 http://www.genome.jp/kegg/pathway.html 9 http://www.pharmgkb.org/ 10 http://www4.wiwiss.fu-berlin.de/bizer/d2r-server/

Extending SMW+ with a Linked Data Integration Framework the semantic wiki) in a uni ed manner. It consists of an integrated chain of Web data access, integration, and storage modules implemented in Java and Scala using the Jena framework. The modules and their tasks are described in the following. 1. Web Data Access Module The basic means to access Linked Data on the Web is to dereference HTTP URIs and to discover additional data sources by traversing RDF links, are realized using the LDspider11 Linked Data crawler. In addition, our LDimporter module can also import data from various RDF dump formats or SPARQL endpoints using data set descriptions based on voiD descriptions [ 1 ] and Semantic Sitemaps [ 4 ]. 2. Vocabulary Mapping Module Di erent Linked Data sources often use di erent vocabularies to represent the same type of information. For instance, both ABA and Uniprot ontologies contain the concept of a gene, but associate di erent relations and properties with it. In order to make Web data understandable to wiki users, the R2R Framework12 is employed as the vocabulary mapping module in order to translate terms from di erent vocabularies into the wiki ontology. 3. Identity Resolution Module Di erent Linked Data sources use di erent URIs to identify the same entity, such as genes or proteins. The framework integrates Silk Server13 as its identity resolution module, which interlinks newly discovered data about entities with data about them that is already de ned in the wiki or other connected data sources. 4. Quality Evaluation Module In order to prefer data from sources known for good quality and to resolve data con icts [ 2 ], the framework includes a data quality evaluation module, the Trust Policy Evaluation Engine (TPEE). 5. Integrated Web Data The repository nally stores the Web data together with provenance information to be used by the application layer. We employ the Named Graphs data model [ 3 ] for representing Web data together with provenance information as an integrated model. 3

Using Linked Data in the Wiki Making the integrated neurogenomics-relevant data available in a semantic wiki empowers users to access it in various ways: { An ontology browser (cf. gure 2) of SMW+ enables the exploration of the integrated data, including representation of provenance information. In this way, correlations between data from di erent sources, as well as data con icts can be identi ed in an interactive manner. { Inline queries (cf. gure 2) and the interactive query interface o er means for ad hoc queries against the set of integrated Linked Data sources. The query results can be organized in wiki pages and complemented with textual descriptions or interpretations. 11 http://code.google.com/p/ldspider/ 12 http://www4.wiwiss.fu-berlin.de/bizer/r2r/ 13 http://www4.wiwiss.fu-berlin.de/bizer/silk/server/

The most prominent bene t of using a semantic wiki environment in this project is the collaborative creation of semantic meta-data via annotations. Rather than just querying integrated Web data, users of the wiki can make use of the data by referring to imported genes or proteins in their own articles. By doing so, further statements about Web data resources are created that are in turn exported as Linked Data. This includes cross data source connections { for instance, the Allen Institute can publish their own data in the wiki and cross-reference it to Uniprot proteins.

Acknowledgement

This work was supported in part by Vulcan Inc. as part of its Project Halo (www.projecthalo.com).

References

1. Alexander , K. , Cyganiak , R. , Hausenblas , M. , Zhao , J. Describing linked datasets . In Proceedings of the 2nd Workshop on Linked Data on the Web (LDOW2009) , 2009 .

2. Bleiholder , J. , Naumann , F. Data fusion . ACM Computing Surveys , 41 ( 1 ):1{ 41 , 2008 .

3. Carroll , J. , Bizer , C. , Hayes , P. , Stickler , P. Named graphs . Journal of Web Semantics , 3 ( 4 ): 247 { 267 , 2005 .

4. Cyganiak , R. , Delbru , R. , Stenzhorn , H. , Tummarello , G. , Decker , S. Semantic sitemaps: E cient and exible access to datasets on the semantic web . In Proceedings of the 5th European Semantic Web Conference (ESWC2008) , 2008 .

5. Jentzsch , A. , Hassanzadeh , O. , Bizer , C. , Andersson , B. , Stephens , S. Enabling tailored therapeutics with linked data . In Proceedings of the 2nd Workshop on Linked Data on the Web (LDOW2009) , 2009 .

6. Krotzsch, M. , Vrandecic , D. , Volkel, M. , Haller , H. , Studer , R. Semantic wikipedia . Journal of Web Semantics , 5 : 251 { 261 , September 2007 .