期刊
INFORMATION SCIENCES
卷 623, 期 -, 页码 40-55出版社
ELSEVIER SCIENCE INC
DOI: 10.1016/j.ins.2022.10.052
关键词
Deep variational inference; Variational auto -encoder; Multi -view document modeling; Hierarchical topic representation; Probability generative model
With the development of the internet, multi-view text documents have become common and research on multi-view document modeling has increased. Traditional single-view document modeling treats each document independently, while multi-view text documents have complex correlations. This study introduces a deep generative model called Hierarchical Variational Auto-Encoder (HVAE) that combines the advantages of probability generative models and deep neural networks. The proposed method successfully captures both global and local topical information in multi-view documents.
With the widespread development of the internet, multi-view text documents have become increasingly common, which has led to extensive research on multi-view text doc-ument modeling. As opposed to traditional single-view document modeling, which treats each document independently and learns each document as a single topic representation, the views of multi-view text documents have complicated correlation relationships that include both the global and local underlying topical information. In this study, we intro-duce a deep generative model for multi-view document modeling known as Hierarchical Variational Auto -Encoder (HVAE), which combines the advantages of the probability gen-erative model for learning interpretable latent information and the deep neural network for efficient parameter inference. Specifically, a set of hierarchical topic representations is learned for each multi-view document to capture the document-level global topical information and view-level local topical information for each view. A two-level hierarchical topic inference network is investigated as the encoder network of HVAE, which is designed using an aligned variational auto-encoder, to learn the hierarchical topic representations. Subsequently, multi-view documents are generated through a two-layered generation net-work, considering both the view-level local and document-level global topic representa-tions. Experiments on three real datasets of different scales for various tasks demonstrate the satisfactory results of the proposed method.(c) 2022 Elsevier Inc. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据