The Norwegian State of Estate (SoE) report includes information about all Norwegian state-owned properties and buildings in the public sector and aims to assist government decision makers to allocate resources more e ectively. A Linked Data based approach is presented here to increase the transparency in the government administration, improve the report generating process and also the report quality. Crossdomain government data originated from the business entity register, the cadastral system, the building accessibility register and the old SoE report are acquired, prepared, cleaned, transformed to Linked Data format and published. The source datasets are then integrated, augmented and interlinked before the results are published as a SPARQL endpoint, used for data visualization and report generation.
The SoE report issued by Statsbygg3 on behalf of the Ministry of Local
Government and Modernisation (KMD) is a governmental white paper, providing a
complete list of state-owned properties and buildings in the Norwegian public
sector. The report has been produced every four years as a result of manual
collection of information from multiple sources. The data collection and
quality control process has historically been resource demanding and error prone
and the result was static and did not re ect the changes after the report was
published. This paper describes the publishing and integration of several
crossdomain government datasets related to state-owned real estates as Linked Open
Data. Sharing the state-owned properties related data in a Linked Data format
enables data reuse, opens up possibilities for using the data in innovative ways,
and helps to increase transparency in the government administration[
In addition, we demonstrate a Web-based application for registration and reporting state-owned properties in Norway. This represents a major improvement compared to the tedious, manual collection of property data. 3 http://www.statsbygg.no/Om-Statsbygg/About-Statsbygg/
Data Sources. Cross-domain government data from di erent open and proprietary sources as listed below are involved in the Linked Data generation process. { The central government organization dataset (a subset of data from the Norwegian Business Entity Register administrated by the Br nn ysund Register Centre4); { The cadastral dataset (a subset of data from the Norwegian Cadastral System5 administrated by the Norwegian Mapping Authority); { The building accessibility dataset from the Building Accessibility Register6 administrated by Statsbygg; { The previous SoE report7 dataset administrated by Statsbygg; { The municipality boundaries dataset8 administrated by the Norwegian Mapping Authority.
The non-geospatial datasets are prepared by dataset providers in tabular format and the geospatial datasets are provided as shape les.
Though the source data comes from the most authoritative sources in the respective domains, it is not always complete and accurate. Data inconsistency between source systems is one of the main challenges in the integration process. Missing values and keys also represent signi cant barriers for the integration process. Such issues are addressed in the Linked Data generation process. Ontology description. Property related data are rich in attributes, with both spatial and temporal characteristics. A common ontology model provides the necessary semantic description of the data. We reuse standard and established ontologies such as DBpedia-owl, DUL, GeoSPARQL and schema.org to represent the data. For example, schema:leiCode, schema:legalName, dbpedia-owl:type, schema:foundingDate and schema:parentOrgnization are used to model organization number, name, type, founding date and parent organization, respectively, in the central government organization dataset.
In addition, we developed the proDataMarket9[
Data cleaning and preparation activities for SoE data included assigning valid cadastral parcel identi ers, unifying null values for attributes with null values, changing decimal formatting and geospatial data conversion. After the source data les were cleaned, they were mapped to the above mentioned ontology and published in DataGraft. Next, the data augmentation was performed with the help of SPARQL CONSTRUCT queries executed on the published data, thus making it possible to infer new data based on known business rules. One example of such a business rule states that a building is owned by the owner or lessor of a cadastral parcel where the building is built upon. This helped to select out the state-owned properties and buildings, calculate the area of real estates, infer the building ownership based on the information about owner or lessor of the belonging cadastral parcel, etc. In addition, the published data is interlinked with several central LOD datasets (DBpedia13, GeoNames14 and Lenka.no15) in order to increase its re-usability to support queries on cross-domain distributed datasets.
The result of the data augmentation and interlinking process is published on DataGraft and is available through a SPARQL endpoint16 under the Norwegian License of Open Data17 (NLOD). SPARQL queries can be run on the SPARQL endpoint to assess the data quality of the SoE report dataset. 3
During the demonstration, we will present the process of generating Linked Data from the SoE report using DataGraft, and a web-based application for registration and reporting state-owned properties in Norway. The demo scenario will cover uploading raw SoE data to the DataGraft platform, data transformation and publication as a Linked Data, and the the Web-based application. The application allows users to con gure visualizations and browse state-owned properties and relevant data on the map created in CartoDB18 (see Figure 1). The stateowned properties absent in the previous SoE report are explicitly marked on the 12 https://datagraft.io/ 13 http://wiki.dbpedia.org/wiktionary-rdf-extraction 14 http://www.geonames.org/ 15 http://data.lenka.no/ 16 https://datagraft.io/prodatamarket_publisher/sparql_endpoints/ norwegian-state-of-estate-report-04693e1f-4060-48c1-8ab9-888a6c95f6d6 SPARQL querying at this endpoint currently works only in Chrome. 17 https://data.norge.no/nlod/en/1.0 18 https://carto.com/ map (using pink colour), helping users to identify data quality issues such as inconsistencies or missing registrations in the source systems.
Acknowledgements This work is partly funded by the EC H2020 project proDataMarket (Grant number: 644497).