Graph-Boosted Active Learning for Multi-source Entity Resolution

Primpeli, Anna; Bizer, Christian

doi:10.1007/978-3-030-88361-4_11

Cited by 8 publications

(3 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There is a notable transition from traditional statistical and machine learning methods toward adopting deep learning techniques, particularly those rooted in transformer architectures and language models Li, 2020;Ye, 2022). This paradigm shift has resulted in a substantial body of research focused on the implementation of end-to-end entity resolution tasks (Konda, 2018;Konda et al, 2016;Mudgal et al, 2018;Dou et al, 2023;Wang et al, 2021;, Primpeli & Bizer, 2021. The entire process is consolidated into a single integrated model in this context.…”

Section: Entrepreneurship and Sustainability Issuesmentioning

confidence: 99%

Information technology for intellectual analysis of item descriptions in e-commerce

Cherednichenko,

Ivashchenko,

Lincényi

et al. 2023

JESI

View full text Add to dashboard Cite

E-commerce is experiencing a robust surge, propelled by the worldwide digital transformation and the mutual advantages accrued by both consumers and merchants. The integration of information technologies has markedly augmented the efficacy of digital enterprise, ushering in novel prospects and shaping innovative business paradigms. Nonetheless, adopting information technology is concomitant with risks, notably concerning safeguarding personal data. This substantiates the significance of research within the domain of artificial intelligence for e-commerce, with particular emphasis on the realm of recommender systems. This paper is dedicated to the discourse surrounding the construction of information technology tailored for processing textual descriptions pertaining to commodities within the e-commerce landscape. Through a qualitative analysis, we elucidate factors that mitigate the risks inherent in unauthorized data access. The cardinal insight discerned is that the apt utilization of product matching technologies empowers the formulation of recommendations devoid of entailing customers' personal data or vendors' proprietary information. A meticulously devised structural model of this information technology is proffered, delineating the principal functional components essential for processing textual data found within electronic trading platforms. Central to our exposition is the exploration of the product comparison predicated on textual depictions. The resolution of this challenge stands to enhance the efficiency of product searches and facilitate product juxtaposition and categorization. The prospective implementation of the propounded information technology, either in its entirety or through its constituent elements, augurs well for sellers, enabling them to improve a pricing strategy and heightened responsiveness to market sales trends. Concurrently, it streamlines the procurement journey for buyers by expediting the identification of requisite goods within the intricate milieu of e-commerce platforms.

show abstract

Section: Entrepreneurship and Sustainability Issuesmentioning

confidence: 99%

Information technology for intellectual analysis of item descriptions in e-commerce

Cherednichenko,

Ivashchenko,

Lincényi

et al. 2023

JESI

View full text Add to dashboard Cite

show abstract

“…For the currently popular ER methods, matching models fall into four main categories: rule-based methods [14; 15], classification models [16; 17; 18], graph-boosted methods [19], and generative models [20]. In this paper, classification-based matching methods are our central focus.…”

Section: Background and Related Workmentioning

confidence: 99%

Transformer-based Denoising Adversarial Variational Entity Resolution

2023

J Intell Inf Syst

View full text Add to dashboard Cite

Entity resolution (ER), precisely identifying different representations of the same real-world entities, is critical for data integration. The ER question has been studied for many years, and many methods have been proposed to solve it. Although deep learning has achieved good performance in ER tasks, there are some challenges regarding manual labeling and model transfer. This paper proposes a novel ER model, Transformerbased Denoising Adversarial Variational Entity Resolution (TdavER). For entity embedding, we develop an unsupervised entity embedding model based on denoising autoencoders and pre-trained language models, which takes corrupted input as training data to motivate the encoder to generate rather stable and robust high-quality entity representations. Furthermore, we propose an unsupervised entity feature transformation model based on adversarial variational autoencoders to remove the constraints on entity representations from training data. This transformation model converts low-level entity embeddings to high-level probability distributions, which are not constrained by the source data and contain deep similarity features. To better implement the feature transformation, we adopt adversarial networks to optimize the variational autoencoder's training process and help it learn the correct posterior distribution. Extensive experiments confirms that the performance of our proposed TdavER is comparable with the current state-of-the-art ER methods and that its entity feature transformation model is transferable.

show abstract

“…It uses human generated rules which can be further improved by supervised methods. For multi source entity matching there exists matching systems like FAMER [30] and ALMSER [28] which in addition uses active learning to reduce the size of the initial alignment.…”

Section: Related Workmentioning

confidence: 99%

DBkWik++ -- Multi Source Matching of Knowledge Graphs

Hertling¹,

Paulheim²

2022

Preprint

View full text Add to dashboard Cite

Large knowledge graphs like DBpedia and YAGO are always based on the same source, i.e., Wikipedia. But there are more wikis that contain information about long-tail entities such as wiki hosting platforms like Fandom. In this paper, we present the approach and analysis of DBkWik++, a fused Knowledge Graph from thousands of wikis. A modified version of the DBpedia framework is applied to each wiki which results in many isolated Knowledge Graphs. With an incremental merge based approach, we reuse one-to-one matching systems to solve the multi source KG matching task. Based on this alignment we create a consolidated knowledge graph with more than 15 million instances.

show abstract

Graph-Boosted Active Learning for Multi-source Entity Resolution

Cited by 8 publications

References 24 publications

Information technology for intellectual analysis of item descriptions in e-commerce

Information technology for intellectual analysis of item descriptions in e-commerce

Transformer-based Denoising Adversarial Variational Entity Resolution

DBkWik++ -- Multi Source Matching of Knowledge Graphs

Contact Info

Product

Resources

About