Tabular open government data search for data spaces based on word embeddings

Abstract

Nowadays, data spaces are envisioned as a prominent mechanism for data sharing, boosting growth and creating value. Open government data providers should be considered as important participants in data space reference infrastructures, since open data portal initiatives are adopted by most of the governments to supply their public sector information. However, open data is mostly published in the form of tabular data such as spreadsheets or CSV files. Therefore, reusing open data in data space is challenging due to the friction that may occur when combining the use of data shared in data spaces and the use of tabular data published in open government portals. To alleviate this situation, tabular open data search engines can be a promising solution. Actually, most open data portals allow reusers to retrieve and federate tabular open data by means of a keyword-based search engine over metadata. Unfortunately, these search engines rely on the (not so often good enough) metadata quality, which must be complete, descriptive, and representative of the content. Moreover, keyword-based search is not always an adequate solution for retrieving open data, since it does not consider their tabular nature and search results can be useless for reusers (e.g., when they attempt to find data useful for extending rows or columns of a given tabular dataset). To overcome these problems, this paper presents an approach that uses word embeddings for tabular open data search based on unionability and joinability. Our approach could be seamlessly integrated in a data space infrastructure. A prototype of our approach has been developed. Finally, both, an intrinsic and an extrinsic evaluation with end users, have been carried out.