In this paper, we describe an application that, starting from statistical classifications represented as Linked Data artefacts, allows for a rich user experience based on cross-classifications navigation. Linked Data and Semantic Web standards are being progressively adopted by statistical organizations. Classifications are an indispensable instrument to disseminate statistical data: without classifications no statistics. An integrated system of activity and product classifications allows the comparability of statistics produced in different statistical domains and countries.
Introduction 2.1
The main functionalities of the Classification Explorer are detailed and shown below.
For some of the uploaded classifications (so far 14), a browsing functionality is available that permits to:
The Classification Explorer also has a search function. Given a code of a classification item or a text to search, the functionality returns all the matching classification items for all the classification that are in the database. The results are grouped by each classification they belong to (see Figure 4).
In addition to the browsing and navigating functionalities illustrated above, the Classification Explorer also has an exporting functionality available. This functionality permits to export and download the displayed list of items in CSV format. 2.2
The Classification Explorer is a single-page browser app that communicates with an RDF triple store. The client is a
(https://nodejs.org/en/),
ReactJS (https://github.com/reactjs) for building user interfaces and Webpack (https://github.com/webpack/webpack), a tool to package, deploy and redeploy applications. The client is based on the React-Redux pattern (http://redux.js.org/) that manages the state of the application. As the application is single-page, the navigation is entirely done on the browser. React-router (https://github.com/reactjs/react-router) takes care of updating the URL so that the user can refresh, resume later or share the page she/he is on. It fetches the data from a remote server, detailed in the next section, using SPARQL 1.1 queries over HTTP. Data are then stored locally and so they are accessed only once.
The server-side is an RDF triple store implemented with Stardog (http://stardog.com/). The data made available within the SemStats challenge have been all uploaded to the triple store. These are RDF data, modelled according to several vocabularies, the principal of which is XKOS. In particular, the classifications are organized in a single database with a named graph for each specific classification version. Figure 5 shows the result of a query for retrieving all the graphs.
The design and implementation of the Classification Explorer was driven by the idea that it is useful to have an integrated access to the various official classifications. Moreover, by exploiting the available correspondence tables, the Classification Explorer has the important advantage of enriching the user experience when accessing classifications by allowing a cross-navigation among them. The application code is available at: https://github.com/UNECE/Classification-Explorer.
When implementing the Classification Explorer, there were some issues related to the lack of harmonized design choices for RDF triples generation. For instance, in the European classifications (NACE and CPA, two versions of each), we had to fix the fact that they used xkos:hasLevels instead of xkos:levels.
The Classification Explorer can be enriched in several ways, namely:
The layout of the HTML pages can be improved. Indeed, so far only a basic layout is adopted. The navigation features can be improved by using URIs for concepts representing items in the correspondence tables instead of strings that prevent from a full navigation.
Enriching the export and download functionality by having bulk downloads in different formats of the available classifications.
i The application and the paper were a joint product of the participants of the Sprint Workshop of the UNECE HLG Linked Statistical Metadata project: Raffaella Aracri, Mauro Bruno, Franck Cotton, Eric Debonnel, Taeke Gjaltema, Dennis Grofils, Hans van Hoof, Olivier Levitt, Enrico Orsini, Andrea Pagano, Jean-Baptiste Rudant, Monica Scannapieco, Romain Tailhurat, Laura Tosco and Luca Valentino.
ii High-Level Group for the Modernisation of Official Statistics (HLG-MOS) was established in 2010 to oversee and coordinate international work relating to statistical modernisation. It promotes standards-based modernisation of official statistics. The aim is to improve the efficiency of statistical production, and the ability to produce outputs that better meet user needs. One of the ways to collaborate and to make progress is through HLG projects in specific areas.