Data Extraction from Structured Formats

The extraction from structured formats, namely, XML and CSV, started at different phases of the project and was carried out by different groups, therefore the used technology slightly varied. The first XML data addressed was the (British) Contracts Finder, [1] for which a standalone XSLT script for transforming all fields to RDF triples was developed in early 2012. Later, however, the main focus was on the European (TED), Czech, Polish, and also U.S. data (to have an extra-European source for comparison).

TED Data

In March 2014 the Publications Office of the EU opened access to the data from TED and ceased to charge licensing fees for data access. Current public notices for the past month are available to download for registered users of the TED portal and also via an FTP server. Archived notices dating back to 2010 can be obtained in monthly data exports. Data is published in 3 formats, including a plain-text one and 2 XML formats.

We created an XSL transformation script to convert the TED data into RDF. Using this XSLT script we performed a bulk extraction of the TED archival data via the Valiant tool[2]) from the LOD2 Stack. In parallel, using the UnifiedViews ETL framework,[3] we set up an automatic, continuously running extraction of the increments in TED data. In the further treatment of the extracted RDF data we focused on deduplication and fusion of business entities participating in the EU public procurement market, in order to provide a more integrated view on the dataset.

Czech Data

We developed an extractor data processing unit[4] for the UnifiedViews ETL framework, which is capable of incremental extraction of data from the Czech public procurement register [5] using its SOAP API. During the time we discussed the possibility of publishing raw open data in bulk with the company running the register. As a result of these discussions we were provided with an XML dump of historical data from the register to be used for research purposes. Combining the historical data dump with the access to current data via the SOAP API we were able to reconstruct the complete dataset of public contracts from the registry converted to RDF.

The second source of Czech public procurement data that we processed was a set of profile feeds of individual contracting authorities. As per the amendments in the Czech public procurement law, public sector bodies involved in public procurement are required to publish their own XML feed of data about public contracts they issue, including both public notices and award information. The set of public contracts that are published on profile feeds is a superset of what is available via the central Czech public procurement registry because the feeds also cover some lower price public contracts, which are not required to be published in the central register. The content of these feeds mostly mirrors the content of the central register, although for individual public contracts it is less comprehensive. While the data from the register is richer and more descriptive, the profile feeds contain information about unsuccessful tenders, which is missing from the register that only reveal information about winning tenders. We deem having data about both successful and unsuccessful tenders as vital in several analytical tasks over public procurement data, which is one of the reasons why we have invested effort into acquiring the data from feeds of contracting authorities. Since early autumn 2013 we have been scraping an HTML list of URLs of profile feeds and periodically convert each feed's XML into RDF using an ETL pipeline developed using the UnifiedViews framework. By using code-based URIs the data is linked to several external datasets. Company identifiers connect it to the Czech business register[6] that we also periodically convert to RDF. Common Procurement Vocabulary (CPV) codes[7] link it to the RDF version of CPV that we produced.

  • [1]
  • [2]
  • [3]
  • [4] extractor
  • [5]
  • [6] register
  • [7] By its definition from codes-cpv en.htm, “CPV establishes a single classification system for public procurement aimed at standardising the references used by contracting authorities and entities to describe the subject of procurement contracts.”
< Prev   CONTENTS   Next >