Hyper-Label-Graph: Modeling Branch-Level Dependencies of Labels for Hierarchical Multi-Label Text Classification

Wenmin Deng, Jing Zhang, Peng Zhang, Yitong Yao, Hui Gao, Yurui Zhang
Proceedings of the 15th Asian Conference on Machine Learning, PMLR 222:279-294, 2024.

Abstract

In the task of Hierarchical Multi-label Text Classification (HTMC), there exist multiple multivariate relations between labels, particularly the semantic dependencies within label branches of the hierarchy. However, existing methods struggle to fully exploit these potential multivariate dependencies since they can only model binary relationships at best. In this paper, we address this limitation by focusing on leveraging semantic dependencies among labels within branches and propose a Hyper-Label-Graph Model (HLGM). Specifically, we first construct a label hypergraph based on the taxonomy hierarchy and utilize a hypergraph attention mechanism to learn branch-level multivariate dependencies among labels. Furthermore, the model employs a label-text fusion module to generate label-level text representations, facilitating the comprehensive integration of semantic features between text and labels. Additionally, we introduce a hierarchical triplet loss to enhance the ability to distinguish labels within the hyperedge structure. We validate the effectiveness of the proposed model on three benchmark datasets, and the experimental results demonstrate that HLGM outperforms competitive GNN-based baselines.

Cite this Paper

BibTeX
@InProceedings{pmlr-v222-deng24b, title = {{Hyper-Label-Graph}: {M}odeling Branch-Level Dependencies of Labels for Hierarchical Multi-Label Text Classification}, author = {Deng, Wenmin and Zhang, Jing and Zhang, Peng and Yao, Yitong and Gao, Hui and Zhang, Yurui}, booktitle = {Proceedings of the 15th Asian Conference on Machine Learning}, pages = {279--294}, year = {2024}, editor = {Yanıkoğlu, Berrin and Buntine, Wray}, volume = {222}, series = {Proceedings of Machine Learning Research}, month = {11--14 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v222/deng24b/deng24b.pdf}, url = {https://proceedings.mlr.press/v222/deng24b.html}, abstract = {In the task of Hierarchical Multi-label Text Classification (HTMC), there exist multiple multivariate relations between labels, particularly the semantic dependencies within label branches of the hierarchy. However, existing methods struggle to fully exploit these potential multivariate dependencies since they can only model binary relationships at best. In this paper, we address this limitation by focusing on leveraging semantic dependencies among labels within branches and propose a Hyper-Label-Graph Model (HLGM). Specifically, we first construct a label hypergraph based on the taxonomy hierarchy and utilize a hypergraph attention mechanism to learn branch-level multivariate dependencies among labels. Furthermore, the model employs a label-text fusion module to generate label-level text representations, facilitating the comprehensive integration of semantic features between text and labels. Additionally, we introduce a hierarchical triplet loss to enhance the ability to distinguish labels within the hyperedge structure. We validate the effectiveness of the proposed model on three benchmark datasets, and the experimental results demonstrate that HLGM outperforms competitive GNN-based baselines.} }
Endnote
%0 Conference Paper %T Hyper-Label-Graph: Modeling Branch-Level Dependencies of Labels for Hierarchical Multi-Label Text Classification %A Wenmin Deng %A Jing Zhang %A Peng Zhang %A Yitong Yao %A Hui Gao %A Yurui Zhang %B Proceedings of the 15th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Berrin Yanıkoğlu %E Wray Buntine %F pmlr-v222-deng24b %I PMLR %P 279--294 %U https://proceedings.mlr.press/v222/deng24b.html %V 222 %X In the task of Hierarchical Multi-label Text Classification (HTMC), there exist multiple multivariate relations between labels, particularly the semantic dependencies within label branches of the hierarchy. However, existing methods struggle to fully exploit these potential multivariate dependencies since they can only model binary relationships at best. In this paper, we address this limitation by focusing on leveraging semantic dependencies among labels within branches and propose a Hyper-Label-Graph Model (HLGM). Specifically, we first construct a label hypergraph based on the taxonomy hierarchy and utilize a hypergraph attention mechanism to learn branch-level multivariate dependencies among labels. Furthermore, the model employs a label-text fusion module to generate label-level text representations, facilitating the comprehensive integration of semantic features between text and labels. Additionally, we introduce a hierarchical triplet loss to enhance the ability to distinguish labels within the hyperedge structure. We validate the effectiveness of the proposed model on three benchmark datasets, and the experimental results demonstrate that HLGM outperforms competitive GNN-based baselines.
APA
Deng, W., Zhang, J., Zhang, P., Yao, Y., Gao, H. & Zhang, Y.. (2024). Hyper-Label-Graph: Modeling Branch-Level Dependencies of Labels for Hierarchical Multi-Label Text Classification. Proceedings of the 15th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 222:279-294 Available from https://proceedings.mlr.press/v222/deng24b.html.

Related Material