Roomba: automatic validation, correction and generation of dataset metadata

Data is being published by both the public and private sectors and covers a diverse set of domains ranging from life sciences to media or government data. An example is the Linked Open Data (LOD) cloud which is potentially a gold mine for organizations and individuals who are trying to leverage external data sources in order to produce more informed business decisions. Considering the significant variation in size, the languages used and the freshness of the data, one realizes that spotting spam datasets or simply finding useful datasets without prior knowledge is increasingly complicated. In this paper, we propose Roomba, a scalable automatic approach for extracting, validating, correcting and generating descriptive linked dataset profiles. While Roomba is generic, we target CKAN-based data portals and we validate our approach against a set of open data portals including the Linked Open Data (LOD) cloud as viewed on the DataHub. The results demonstrate that the general state of various datasets and groups, including the LOD cloud group, needs more attention as most of the datasets suffer from bad quality metadata and lack some informative metrics that are required to facilitate dataset search.