Automatic Academic Paper Rating Based on Modularized Hierarchical Attention Network

Kai Kang, Huaping Zhang*, Yugang Li, Xi Luo, Silamu Wushour

*此作品的通讯作者

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

Automatic academic paper rating (AAPR) remains a difficult but useful task to automatically predict whether to accept or reject a paper. Having found more task-specific structure features of academic papers, we present a modularized hierarchical attention network (MHAN) to predict paper quality. MHAN uses a three-level hierarchical attention network to shorten the sequence for each level. In the network, the modularized parameter distinguishes the semantics of functional chapters. And a label-smoothing mechanism is used as a loss function to avoid inappropriate labeling. Compared with MHCNN and plain HAN on an AAPR dataset, MHAN achieves a state-of-the-art accuracy of 65.33%. Ablation experiments show that the proposed methods are effective.

源语言英语
主期刊名Natural Language Processing and Chinese Computing - 11th CCF International Conference, NLPCC 2022, Proceedings
编辑Wei Lu, Shujian Huang, Yu Hong, Xiabing Zhou
出版商Springer Science and Business Media Deutschland GmbH
669-681
页数13
ISBN(印刷版)9783031171192
DOI
出版状态已出版 - 2022
活动11th CCF International Conference on Natural Language Processing and Chinese Computing, NLPCC 2022 - Guilin, 中国
期限: 24 9月 202225 9月 2022

出版系列

姓名Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
13551 LNAI
ISSN(印刷版)0302-9743
ISSN(电子版)1611-3349

会议

会议11th CCF International Conference on Natural Language Processing and Chinese Computing, NLPCC 2022
国家/地区中国
Guilin
时期24/09/2225/09/22

指纹

探究 'Automatic Academic Paper Rating Based on Modularized Hierarchical Attention Network' 的科研主题。它们共同构成独一无二的指纹。

引用此