Wombat – A Generalization Approach for Automatic Link Discovery

Sherif, Mohamed Ahmed; Ngomo, Axel-Cyrille Ngonga; Lehmann, Jens

doi:10.1007/978-3-319-58068-5_7

Cited by 30 publications

(27 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They assume that if two entities are very similar, they are likely the same. Hence, such specifications may generate links through (adapted from [44]):…”

Section: Motivationsmentioning

confidence: 99%

See 1 more Smart Citation

Link key candidate extraction with relational concept analysis

Atencia

Jérôme

Euzenat

et al. 2020

Discrete Applied Mathematics

View full text Add to dashboard Cite

Linked data aims at publishing data expressed in RDF (Resource Description Framework) at the scale of the worldwide web. These datasets interoperate by publishing links which identify individuals across heterogeneous datasets. Such links may be found by using a generalisation of keys in databases, called link keys, which apply across datasets. They specify the pairs of properties to compare for linking individuals belonging to different classes of the datasets. Here, we show how to recast the proposed link key extraction techniques for RDF datasets in the framework of formal concept analysis. We define a formal context, where objects are pairs of resources and attributes are pairs of properties, and show that formal concepts correspond to link key candidates. We extend this characterisation to the full RDF model including non functional properties and interdependent link keys. We show how to use relational concept analysis for dealing with cyclic dependencies across classes and hence link keys. Finally, we discuss an implementation of this framework.stating that whenever an instance of the class Livre has the same values for the property auteur as an instance of the class Book has for the property creator and they share at least one value for their properties titre and title, then they denote the same entity. The notion of link key used in this paper generalises the definition introduced in [5] because the body of the rule includes two sets: the set of property pairs for which instances have to share all the values and the set of property pairs for which instances have to share at least one value.Such a link key may depend on other ones. For instance, properties auteur and creator may have values in the Écrivain and Writer classes respectively. Identifying their values will then resort to another link key:{⟨prénom,firstname⟩}{⟨nom,lastname⟩} linkkey ⟨Écrivain,Writer⟩ This situation would be even more intricate if Écrivain and Writer were instead identified from the values of their properties ouvrages and hasWritten refering to instances of Livre and Book (which have a chance to be more accurate). We would then face interdependent link keys.The problem considered here is the extraction of such link keys from RDF data. We have already proposed an algorithm for extracting some types of link keys [5]. This method may be decomposed in two distinct steps:(1) identifying link key candidates, i.e. sets of property pairs that would generate at least one link if used as a link key and that would be maximal for at least one generated link, followed by (2) selecting the best link key candidates according to quality measures. A method for discovering functional link keys in relational databases based on Formal Concept Analysis (FCA [23]) is detailed in [6].Globally extracting a set of link keys across several RDF data sources raises several issues, as link keys differ from database keys in various aspects: (a) they relax two constraints of the relational model, namely, that attributes are functional (RDF properties may have s...

show abstract

“…They assume that if two entities are very similar, they are likely the same. Hence, such specifications may generate links through (adapted from [44]):…”

Section: Motivationsmentioning

confidence: 99%

“…Various methods have been designed for extracting numerical specifications based on spatial techniques [43], probabilities [46], or genetic programming and active learning [36]. Such numerical specifications may be extracted by using machine learning [27,44]. A larger selection of methods is available in [34].…”

Section: Motivationsmentioning

confidence: 99%

Link key candidate extraction with relational concept analysis

Atencia

Jérôme

Euzenat

et al. 2020

Discrete Applied Mathematics

View full text Add to dashboard Cite

show abstract

“…The authors of [27] report a comprehensive survey of Link Discovery frameworks, which shows that modern framework such as Silk [28,4], LIMES [29], EAGLES [5] combine manually defined match rules with genetic programming and/or active learning approaches to automatize the configuration process. A different approach is proposed by WOMBAT [30], which relies on an iterative search process based on an upward refinement operator. WOMBAT learns to combine atomic link specifications into complex link specifications to optimize the F-score using only positive examples.…”

Section: Related Workmentioning

confidence: 99%

“…We re-used some classes and properties from the following ontologies: dul, 24 schema, 25 dc, 26 lode, 27 geo, 28 transit, 29 and topo. 30 A few additional classes and properties have been created to describe travel distances: we defined origin, distance, travel time, the nearest metro station and bike station. Details are available in [56].…”

Section: Cixty Knowledge Base Generationmentioning

confidence: 99%

STEM: Stacked threshold-based entity matching for knowledge base generation

Palumbo

Rizzo

Troncy

2018

View full text Add to dashboard Cite

One of the major issues encountered in the generation of knowledge bases is the integration of data coming from a collection of heterogeneous data sources. A key essential task when integrating data instances is the entity matching. Entity matching is based on the definition of a similarity measure among entities and on the classification of the entity pair as a match if the similarity exceeds a certain threshold. This parameter introduces a trade-off between the precision and the recall of the algorithm, as higher values of the threshold lead to higher precision and lower recall, and lower values lead to higher recall and lower precision. In this paper, we propose a stacking approach for threshold-based classifiers. It runs several instances of classifiers corresponding to different thresholds and use their predictions as a feature vector for a supervised learner. We show that this approach is able to break the trade-off between the precision and recall of the algorithm, increasing both at the same time and enhancing the overall performance of the algorithm. We also show that this hybrid approach performs better and is less dependent on the amount of available training data with respect to a supervised learning approach that directly uses properties' similarity values. In order to test the generality of the claim, we have run experimental tests using two different threshold-based classifiers on two different data sets. Finally, we show a concrete use case describing the implementation of the proposed approach in the generation of the 3cixty Nice knowledge base.

show abstract

“…This may also be useful if the data related to a particular class is the result of aggregating different sources that use different properties: there may be several different ways to generate links. Thus, instead of selecting one single best link key candidate, it could be worth selecting the best combination of link key candidates, as it is already done for other link specifications [Sherif et al 2017]. This paper addresses the specific problem of extracting boolean combinations of link key candidates from two RDF data sets.…”

Section: Introductionmentioning

confidence: 99%

Several Link Keys are Better than One, or Extracting Disjunctions of Link Key Candidates

Atencia

Jérôme

Euzenat

2019

Proceedings of the 10th International Conference on Knowledge Capture

View full text Add to dashboard Cite

Link keys express conditions under which instances of two classes of different RDF data sets may be considered as equal. As such, they can be used for data interlinking. There exist algorithms to extract link key candidates from RDF data sets and different measures have been defined to evaluate the quality of link key candidates individually. For certain data sets, however, it may be necessary to use more than one link key on a pair of classes to retrieve a more complete set of links. To this end, in this paper, we define disjunction of link keys, propose strategies to extract disjunctions of link key candidates from RDF data, and apply existing quality measures to evaluate them. We also report on experiments with these strategies.

show abstract

Wombat – A Generalization Approach for Automatic Link Discovery

Cited by 30 publications

References 18 publications

Link key candidate extraction with relational concept analysis

Link key candidate extraction with relational concept analysis

STEM: Stacked threshold-based entity matching for knowledge base generation

Several Link Keys are Better than One, or Extracting Disjunctions of Link Key Candidates

Contact Info

Product

Resources

About