Assessing linkset quality for complementing third-party datasets

Abstract

Linked data best practices are getting extremely popular: various companies and public institutions have started taking advantage of linked data principles for exposing their datasets, and for relating their datasets to those served by third parties. Such enthusiasm is due to the linked data promise of evolving into a Global Data Space. Linksets are sets of links relating datasets and they surely play a fundamental role in this promise. However, a stable and wellaccepted notion of linkset quality has not been yet defined. This paper contributes to overcome this lack by proposing a linkset quality measure. Among the different quality dimensions that can be addressed, the proposed measure focuses on completeness. The paper formally defines novel scoring functions and proposes an interpretation of these functions when maintaining and complementing third party datasets.

Quality assessment for Linked Data: A Survey

“…The two types of completeness defined in Mendes et al can be mapped to the two categories (i) Schema completeness and (iii) Population completeness provided by Fürber et al Additionally, we introduce the category interlinking completeness, which refers to the degree to which instances in the dataset are interlinked [25]. Albertoni et al [2] define interlinking completeness as "linkset completeness as the degree to which links in the linksets are not missing." The other articles [17,50,53] provide metrics for this dimension.…”

“…In terms of LD, completeness comprises of the following aspects: (i) Schema completeness, the degree to which the classes and properties of an ontology are represented, thus can be called "ontology completeness", (ii) Property completeness, measure of the missing values for a specific property, (iii) Population completeness is the percentage of all real-world objects of a particular type that are represented in the datasets and (iv) Interlinking completeness, which has to be considered especially in LD, refers to the degree to which instances in the dataset are interlinked. [25,53], (ii) calculating percentage of mappable types in a datasets that have not yet been considered in the linksets when assuming an alignment among types [2] It should be noted that in this case, users should assume a closed-world-assumption where a gold standard dataset is available and can be used to compare against the converted dataset.…”

“…Additionally, it can be seen that 12 approaches assess data on both a triple and graph level [10,11,14,18,19,22,36,40,48,50,53,55], two approaches assess data both at graph and dataset level [20,25] and five approaches assess data at triple, graph and dataset levels [8,23,30,31,65]. There are seven approaches that apply the assessment only at triple level [1,2,15,17,27,45,64] and four approaches that only apply the assessment at the graph level [21,24,32,59].…”

Methods for Intrinsic Evaluation of Links in the Web of Data

“…They examined the number of valid and dead links (i. e. in their work, links with an o that cannot be described in the target distribution), as well as the number of namespaces in link distributions and data sets. Albertoni et al [2,1] analysed the completeness of the interlinking of pairs of data sets and the extent to which data sets become more multilingual thanks to the links. These methods fail in stating the extent to which links add value to the source data set in terms of the principles that we mention in this paper.…”

“…Neto et al [10] focused on the analysis of dead links in schema and entity link triples published in the Web of Data. While these studies, together with the findings provided by smaller evaluations of other link assessment methods focusing on (i) (e. g. [5] and other quality dimensions like completeness [2] provide a characterization of existing links, they do not allow for assessing how many new things might be made discoverable thanks to the links (ii).…”