Targeted Sentiment Classification with Attentional Encoder Network

被引:99
作者
Song, Youwei [1 ]
Wang, Jiahai [1 ]
Jiang, Tao [1 ]
Liu, Zhiyue [1 ]
Rao, Yanghui [1 ]
机构
[1] Sun Yat Sen Univ, Sch Data & Comp Sci, Guangzhou, Peoples R China
来源
ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING - ICANN 2019: TEXT AND TIME SERIES, PT IV | 2019年 / 11730卷
基金
中国国家自然科学基金;
关键词
Target-dependent sentiment classification; Sentiment classification; Sentiment analysis; Attention mechanism;
D O I
10.1007/978-3-030-30490-4_9
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Targeted sentiment classification aims at determining the sentimental tendency towards specific targets. Most of the previous approaches model context and target words with RNN and attention. However, RNNs are difficult to parallelize and truncated backpropagation through time brings difficulty in remembering long-term patterns. To address this issue, this paper proposes an Attentional Encoder Network (AEN) which eschews recurrence and employs attention based encoders for the modeling between context and target. We raise the label unreliability issue and introduce label smoothing regularization. We also apply pre-trained BERT to this task and obtain new state-ofthe-art results. Experiments and analysis demonstrate the effectiveness and lightweight of our model.
引用
收藏
页码:93 / 103
页数:11
相关论文
共 20 条
[1]  
[Anonymous], 2009, EACL 2009
[2]  
[Anonymous], 2014, 3 INT C LEARN REPR
[3]  
Devlin J., 2018, ARXIV
[4]  
Ding X., 2008, P 2008 INT C WEB SEA, P231, DOI [10.1145/1341531.1341561, DOI 10.1145/1341531.1341561]
[5]  
Dong L, 2014, PROCEEDINGS OF THE 52ND ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, VOL 2, P49
[6]  
Glorot X., 2010, P 13 INT C ART INT S, P249
[7]  
Jiang Long, 2011, P ACL, P151
[8]  
Kingma DP, 2014, ARXIV
[9]  
Kiritchenko S., 2014, P 8 INT WORKSH SEM E, P437, DOI DOI 10.3115/V1/S14-2076
[10]  
Li L, 2016, IEEE IND ELEC, P3294, DOI 10.1109/IECON.2016.7793303