Spotting Suspicious Academic Citations Using Self-Learning Graph Transformers

The study introduces a novel approach to identify potential citation manipulation within academic papers. This method utilizes perturbations of a deep embedding model, integrating Graph-Masked Autoencoders to merge textual information with evidence of graph connectivity. Consequently, it yields a mo...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Mathematics (Basel) Ročník 12; číslo 6; s. 814
Hlavní autoři: Avros, Renata, Haim, Mor Ben, Madar, Almog, Ravve, Elena, Volkovich, Zeev
Médium: Journal Article
Jazyk:angličtina
Vydáno: Basel MDPI AG 01.03.2024
Témata:
Analysis
Citations
Computational linguistics
Credibility
Design
Electric transformers
Graph representations
Graph theory
graph-masked autoencoders
Language processing
manipulated citations
Natural language interfaces
network perturbation
Perturbation
ISSN:2227-7390, 2227-7390
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:The study introduces a novel approach to identify potential citation manipulation within academic papers. This method utilizes perturbations of a deep embedding model, integrating Graph-Masked Autoencoders to merge textual information with evidence of graph connectivity. Consequently, it yields a more intricate model of citation distribution. By training a deep network with partial data and reconstructing masked connections, the approach capitalizes on the inherent characteristics of central connections amidst network perturbations. It demonstrates its ability to pinpoint trustworthy citations within the analyzed dataset through comprehensive quantitative evaluations. Additionally, it raises concerns regarding the reliability of specific references, which may be subject to manipulation.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2227-7390
2227-7390
DOI:10.3390/math12060814