Joint Entity Linking for Web Tables with Hybrid Semantic Matching

Xie, Jie; Lu, Yuhai; Cao, Cong; Li, Zhenzhen; Guan, Yangyang; Liu, Yanbing

doi:10.1007/978-3-030-50417-5_46

Cited by 4 publications

(3 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table 10 shows a comparison of the qualitative indicators with the closest works [26][27][28][29][30][31][32][33][35][36][37][40][41][42][43][44][45][47][48][49][50]. We have grouped all existing methods into two classes: heuristic methods and deep-learning-based methods (e.g., using Deep Neural Networks (DNN) or BERT models).…”

Section: Resultsmentioning

confidence: 99%

“…It should also be noted that many solutions seemed to specialize in one particular STI task. For example, [42] proposes a new hybrid semantic matching model called JHSTabEL for entity linking in a table. A feature of this model is the collection of local semantic information between text mentions in a table and candidate entities from the perspective of different semantic aspects.…”

Section: Related Workmentioning

confidence: 99%

“…An exception is the work [58], where the table annotation process uses concepts from three knowledge graphs linked by "sameAs" relationships; • Some approaches are aimed at solving only one STI task and do not provide a comprehensive solution. For example, JHSTabEL [42] is focused only on the CEA task, while ColNet [43] and Sherlock [44] Modern approaches based on deep machine learning (e.g., Sato [45], Doduo [47], TURL [48]) are applicable only to a certain set of tables belonging to the same domain, i.e., tables for annotation and tables on which training was carried out should be similar. Moreover, such approaches represent a limited number of semantic types (classes) for annotation (only those types for which classifiers have been trained).…”

mentioning

confidence: 99%

See 2 more Smart Citations

Knowledge Graph Engineering Based on Semantic Annotation of Tables

Dorodnykh,

Yurin

2023

Computation

View full text Add to dashboard Cite

A table is a convenient way to store, structure, and present data. Tables are an attractive knowledge source in various applications, including knowledge graph engineering. However, a lack of understanding of the semantic structure and meaning of their content may reduce the effectiveness of this process. Hence, the restoration of tabular semantics and the development of knowledge graphs based on semantically annotated tabular data are highly relevant tasks that have attracted a lot of attention in recent years. We propose a hybrid approach using heuristics and machine learning methods for the semantic annotation of relational tabular data and knowledge graph populations with specific entities extracted from the annotated tables. This paper discusses the main stages of the approach, its implementation, and performance testing. We also consider three case studies for the development of domain-specific knowledge graphs in the fields of industrial safety inspection, labor market analysis, and university activities. The evaluation results revealed that the application of our approach can be considered the initial stage for the rapid filling of domain-specific knowledge graphs based on tabular data.

show abstract