Pay-as-you-go Configuration of Entity Resolution

Abstract

Abstract. Entity resolution, which seeks to identify records that represent the same entity, is an important step in many data integration and data cleaning applications. However, entity resolution is challenging both in terms of scalability (all-against-all comparisons are computationally impractical) and result quality (syntactic evidence on record equivalence is often equivocal). As a result, end-to-end entity resolution proposals involve several stages, including blocking to efficiently identify candidate duplicates, detailed comparison to refine the conclusions from blocking, and clustering to identify the sets of records that may represent the same entity. However, the quality of the result is often crucially dependent on configuration parameters in all of these stages, for which it may be difficult for a human expert to provide suitable values. This paper describes an approach in which a complete entity resolution process is optimized, on the basis of feedback (such as might be obtained from crowds) on candidate duplicates. Given such feedback, an evolutionary search of the space of configuration parameters is carried out, with a view to maximizing the fitness of the resulting clusters. The approach is payas-you-go in that more feedback can be expected to give rise to better outcomes. An empirical evaluation shows that the co-optimization of the different stages in entity resolution can yield significant improvements over default parameters, even with small amounts of feedback.

Multi-Source Spatial Entity Linkage

“…The entity resolution problem has been referred in the literature with multiple terms including deduplication, entity linkage, and entity matching [4,15]. Entity resolution has been used in various fields such as matching profiles in social networks [2], bioinformatics data [3], biomedical data [16], publication data [4,5], genealogical data [6], product data [4,5], etc. The attributes of the entities are compared, and a similarity value is assigned.…”

“…Other approaches decide the include the uncertainty of a match into the decision [19]. Finally, matching the entities can also be based on the feedback of an oracle [4,5] or of a user [5].…”

“…(3) to rank the pairs by their similarity, we propose a flexible technique (SkyRank) that is based on the concept of Pareto optimality; (4) to label the pairs, we propose the SkyEx-* family of algorithms that considers the ranking order of the pairs and fixes a cut-off level to separate the classes; (5) we introduce two threshold-based algorithms: SkyEx-F that uses the F-measure to separate the classes, and SkyEx-FES, an optimized version of SkyEx-F, which provides a theoretical guarantee to prune 73% of the skyline explorations of SkyEx-F; (6) we propose SkyEx-D, a novel algorithm that is fully unsupervised and parameter-free to separate the classes.…”

Feedback driven improvement of data preparation pipelines

“…In pay-as-you-go data preparation, following from the vision of dataspaces [12], a best-effort integration is produced through automated bootstrapping, which is refined in the light of user feedback. There have been many proposals for pay-as-you-go data preparation components, for example relating to data extraction [10], matching [18,36], mapping [5,6,30] and entity resolution [14,25]. Such proposals have addressed issues such as targeting the most appropriate feedback [10,28,35], and accommodating unreliable feedback, in particular in the context of crowdsourcing [9,23].…”

“…This may be the closest work to ours, though the focus is on a single data integration step, for which custom feedback has been obtained. Also for entity resolution, feedback on matching pairs is used in [25] as part of a single optimisation step that configures together blocking, detailed comparison and clustering. This contrasts with the current paper in focusing different aspects of the same integration step.…”