Transductive Text Classification With Concept Bipartite Graphs and LLM-Assisted Labeling
IEEE Access · 2026 · 14 · 63698-63715
DOI: 10.1109/access.2026.3684441 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Vitor Vasconcelos de Oliveira und 2 weiteren Autor:innen. Die Arbeit nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Brazilian National Council for Scientific and Technological Development, (Postdoctoral Fellowship Abroad), Coordination for the Improvement of Higher Education Personnel (CAPES), Finance Code 001, and by the São Paulo Research Foundation.
Abstract (Originalsprache)
Graph-based text classification has become an effective paradigm for modeling corpus-level structure, particularly under limited supervision. However, most existing approaches rely on document–word representations and do not jointly address semantic abstraction, connectivity robustness, and supervision scarcity in transductive settings. We investigate transductive text classification in low-label regimes using corpus-level document–concept bipartite graphs, where documents are connected to automatically extracted keyphrases represented by contextual embeddings. This concept-level representation replaces traditional document–word graphs with a more compact and semantically expressive structure. To ensure global information propagation, we introduce a $k$ -nearest concept linking mechanism that enforces graph connectivity and prevents isolated subgraphs. Graph neural networks (GCN and GAT) are applied to this structure, and LLM-assisted labeling is incorporated as weak supervision by generating pseudo-labels with a resource-efficient language model and combining them with limited human annotations. Multilevel bipartite coarsening is used as a complementary strategy to reduce graph size and examine the accuracy–efficiency trade-off. Experiments on 12 datasets show that attention-based message passing is generally more effective than standard convolutions on document–concept graphs, that $k$ -NN concept linking enhances structural robustness, and that LLM-generated labels provide the greatest benefit in extremely low-label settings, with diminishing returns as human supervision increases.
Förderung
- Brazilian National Council for Scientific and Technological Development
- (Postdoctoral Fellowship Abroad)
- Coordination for the Improvement of Higher Education Personnel (CAPES), Finance Code 001, and by the São Paulo Research Foundation
Zitieren
APA
Oliveira, V. V. D., Valejo, A. D. B., & Faleiros, T. D. P. (2026). Transductive Text Classification With Concept Bipartite Graphs and LLM-Assisted Labeling. IEEE Access, 14, 63698-63715. https://doi.org/10.1109/access.2026.3684441
Vancouver
Oliveira VVD, Valejo ADB, Faleiros TDP. Transductive Text Classification With Concept Bipartite Graphs and LLM-Assisted Labeling. IEEE Access. 2026;14:63698-63715. doi:10.1109/access.2026.3684441
BibTeX
@article{Oliveira2026,
author = {Oliveira, Vitor Vasconcelos de and Valejo, Alan Demérius Baria and Faleiros, Thiago de Paulo},
title = {Transductive Text Classification With Concept Bipartite Graphs and LLM-Assisted Labeling},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {63698--63715},
doi = {10.1109/access.2026.3684441}
}
EUVIMED