Conclusions and Future Work
Opening up governmental data requires two elements: discoverability – where more data portals are available, harvesting is used to gather data; and quality machine readable data – where data can be trusted. LOD2 tools support development of these functionalities.
We have addressed the general challenge where data currently being published is in formats and presentation forms intended for direct consumption by humans, but with limited possibilities for automated processing or analysis conducted by a custom software. As one of the important issues for data discoverability we have identiﬁed the need for providing data catalogues. CKAN is a reference solution in this area but it requires further extensions. Working implementations include: publicdata.eu and open-data.europa.eu. The quality of data has been demonstrated by providing Polish and Serbian data ﬁrst automatically converted and then carefully improved.
One of the missing features is cross-language searching. Although the user interface can be used in multiple languages, metadata for datasets can of course be read only in the language in which it was input. A limitation of search on PublicData.eu is that as the source catalogues are naturally in diﬀerent languages, a single search term in any given language will usually not ﬁnd all the relevant datasets.
For wider adoption of CKAN we also need better metadata management. The current harvesting arrangement does not preserve all the original metadata. In particular, even where the harvest source is CKAN-based and datasets have identiﬁable publishing departments, this information is not preserved in the record on PublicData.eu. Adding this information to the harvesting system would enable users on PublicData.eu to 'follow' individual departments and see dashboard notiﬁcations when departments published new or updated data. An example of an alternative harvesting process (built on top of LOD2 stack tools) that preserves the original metadata is available at data.opendatasupport.eu. Several other tools, not mentioned in the chapter, have been particularly useful for making data accessible for machines: Virtuoso, Ontowiki (with CubeViz plug-in), SILK, Open Reﬁne and PoolParty. Various datasets have been elaborated in detail manually, particularly those using the RDF Data Cube vocabulary. Some examples include: national accounts, foreign trade, energy-related, and public procurement data. We have increased the openness of the data by preparing respective vocabularies and providing linking to other data sources available on the Web.
A signiﬁcant amount of time was absorbed by data quality issues. Even though data was available in 'machine processable' XML, it were users who entered incorrect data. These are, however, typical problems of integration projects and should not under any circumstances be considered to be related to the linked data paradigm. On the contrary, applying tools that we had at our disposal allowed to spot quality problems even faster than we would have been able to otherwise.
Open Access. This chapter is distributed under the terms of the Creative Commons Attribution Noncommercial License, which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
1. Berners-Lee, T.: Relational databases on the semantic web, 09 1998. Design Issues. w3.org/DesignIssues/RDB-RDF.html
2. Cyganiak, R., Reynolds, D., Tennison, J.: The RDF Data Cube vocabulary. Technical report, W3C (2013)
3. Ermilov, I., Auer, S., Stadler, C.: Csv2rdf: user-driven csv to rdf mass conversion framework. In: ISEM '13, 04–06 September 2013, Graz, Austria (2013)
4. Kro¨tzsch, M., Vrandecic, D., V¨olkel, M., Haller, H., Studer, R.: Semantic Wikipedia. J. Web Semant. 5, 251–261 (2007)