University of Bahrain
Scientific Journals

Hybrid Approach to Instance Matching

Show simple item record B, Vijaya Gharpure, Prachi 2024-03-25T15:59:10Z 2024-03-25T15:59:10Z 2024-03-23
dc.identifier.issn 2210-142X
dc.description.abstract The proliferation of web data and the advent of extensive knowledge graphs have led to the creation of vast volumes of disconnected data, resulting in data silos. Integrating and sharing web data across diverse domains require that equivalent instances across different sources are correctly identified. Instance matching, often referred to as Entity Resolution, encompasses the task of determining whether two instances correspond to same resource or entity. This process poses significant challenges, particularly in distinguishing between identical entities and those with similar attributes. Candidate generation has a pivotal role in facilitating appropriate comparisons between entities across disparate datasets. This paper employs an inverted index based approach to identify candidates for the matching task and a query likelihood model based selection to further reduce the candidate set. This paper proposes a novel system architecture employing hybrid ensemble classifiers and a methodology for identifying equivalent instances despite the challenges posed by diverse data representations in instance matching. Through experimental evaluation on real-world datasets, we demonstrate that our hybrid ensemble learning approach consistently outperforms standalone matchers in terms of accuracy and F1score. A comprehensive literature review on instance matching, discussing practical considerations and challenges for data interlinking is also presented here. Future research directions aimed at contributing to seamless data integration and knowledge sharing across disparate domains are also outlined. en_US
dc.language.iso en en_US
dc.publisher University of Bahrain en_US
dc.subject Instance matching, Semantic Web, Entity Resolution, Candidate Generation, Ensemble Classification. en_US
dc.title Hybrid Approach to Instance Matching en_US
dc.volume 16 en_US
dc.issue 1 en_US
dc.pagestart 1 en_US
dc.pageend 12 en_US
dc.contributor.authorcountry India en_US
dc.contributor.authorcountry India en_US
dc.contributor.authoraffiliation Thadomal Shahani Engineering College en_US
dc.contributor.authoraffiliation Mumbai University en_US
dc.source.title International Journal of Computing and Digital Systems en_US
dc.abbreviatedsourcetitle IJCDS en_US

Files in this item

This item appears in the following Issue(s)

Show simple item record

All Journals

Advanced Search


Administrator Account