KANG Gang,WU Sijiu,FANG Rui.Extraction of Opinion Terms and Opinion Targets based on Capsule Feature Aggregation[J].Journal of Chengdu University of Information Technology,2020,35(05):524-530.[doi:10.16836/j.cnki.jcuit.2020.05.008]
基于胶囊特征聚合的评价词和评价对象抽取
- Title:
- Extraction of Opinion Terms and Opinion Targets based on Capsule Feature Aggregation
- 文章编号:
- 2096-1618(2020)05-0524-07
- Keywords:
- capsule network; dynamic routing algorithm; opinion targets; opinion terms; IndGRU; sequence labeling
- 分类号:
- TP391.1
- 文献标志码:
- A
- 摘要:
- 为提高传统门控制单元(GRU)的特征抽取能力,受独立循环神经网络(independently recurrent neural network, IndRNN)的启发,提出独立门控制单元(independently gate recurrent unit,IndGRU),实验结果表明IndGRU在3个数据集上的特征提取能力优于传统GRU和长短期记忆网络(LSTM),证明了IndGRU的有效性。针对传统评价词和评价对象抽取方法不能很好地利用抽象特征之间的关联关系问题,提出一种基于胶囊特征聚合的评价词和评价对象抽取模型,模型使用多个双向IndGRU并行提取文本上下文信息,构造胶囊特征,使用动态路由算法利用特征间关系实现胶囊特征的聚合,最后使用条件随机场(CRF)完成序列标注。模型在3个基准数据集上取得了比目前的先进方法较好或相当的效果,证明了模型的有效性。
- Abstract:
- In order to improve the feature extraction ability of Gate Recurrent Unit, inspired by the work of Independently Recurrent Neural Network(IndRNN), Independently Gate Recurrent Unit(IndGRU)is proposed. The experimental results show that the ability of feature extraction of IndGRU is superior to the traditional GRU and LSTM on three benchmark datasets, which prove the IndGRU is effective. Due to that the traditional model for extraction of opinion targets and opinion terms can not make good use of correlation between abstract features, an opinion targets and opinion terms extraction model based on capsule feature aggregation is proposed. This model extracts different context information features by using multiple bidirectional IndGRU in parallel, capsule feature is constructed based on these extracted features, and relationship between features is used by the dynamic routing algorithm to aggregate the capsule features. finally, the aggregated features were used in Conditional Random Field to complete the sequence labeling. The model has achieved performance better than or comparable with the current advanced methods on three benchmark data sets, which proves the effectiveness of this model.
参考文献/References:
[1] Dai H,Song Y.Neuralaspect and opinion term extraction with mined rules as weak supervision[C].Proceedings of the 57th Conference of the Association for Computational Linguistics.Stroudsburg,PA:Association for Computational Linguistics,2019:5268-5277.
[2] Qiu G,Liu B,Bu J,et al.Opinion wordexpansion and target extraction through double propagation[J].Computational Linguistics,2011,37(1):9-27.
[3] Wang W,Pan S J,Dahlmeier D,et al.Recursive neural conditional random fields for aspect-based sentiment analysis[C].Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing,Stroudsburg,PA:ACL,2016:616-626.
[4] Li S,Li W,Cook C,et al.Independently recurrent neural network(indrnn):building a longer and deeper rnn[C].2018 IEEE Conference on Computer Vision and Pattern Recognition.Los Alamitos,CA:IEEE Computer Society,2018:5457-5466.
[5] Hinton G E,Krizhevsky A,Wang S D.Transforming auto-encoders[C].21st International Conference on Artificial Neural Networks.Berlin,Germany:Springer,2011:44-51.
[6] Sabour S,Frosst N,Hinton G E.Dynamic routing between capsules[C].Advances in Neural Information Processing Systems 30:Annual Conference on Neural Information Processing Systems 2017.California,USA:NIPS Proceeding,2017:3856-3866.
[7] 郑毅.时间序列数据的胶囊式LSTM特征提取算法研究[D].武汉:华中师范大学,2018.
[8] Xu L,Liu K,Lai S,et al.Walk and learn:a two-stage approach for opinion words and opinion targets co-extraction[C].Proceedings of the 22nd International Conference on World Wide Web.Rio de Janeiro,Brazil:International World Wide Web Conferences Steering Committee,2013: 95-96.
[9] Liu Qian,Gao Zhiqing,Liu Bing,et al.Automated rule selection for aspect extraction in opinion mining[C].Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence.Menlo Park,CA:AAAI Press,2015:1291-1297.
[10] Niklas Jakob,Iryna Gurevych.2010.Extracting opinion targets in a single and cross-domain setting with conditional random fields[C].Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing.Stroudsburg,PA:Association for Computational Linguistics,2010:1035-1045.
[11] Jin W,Ho H H,Srihari R K.A novel lexicalized HMM-based learning framework for web opinion mining[C].Proceedings of the 26th annual international conference on machine learning.New York:ACM,2009:465-472.
[12] Liu P,Joty S,Meng H.Fine-grained opinion mining with recurrent neural networks and word embeddings[C].Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing.Stroudsburg,PA:Association for Computational Linguistics,2015:1433-1443.
[13] Zhao Y,Qin B,Liu T.Encoding syntactic representations with a neural network for sentiment collocation extraction[J].Science China Information Sciences,2017,60(11):3-14.
[14] Wu C,Wu F,Wu S,et al.A hybrid unsupervised method for aspect term and opinion target extraction[J].Knowledge-Based Systems,2018,148:66-73.
[15] Ratinov L,Roth D.Design challenges and misconceptions in named entity recognition[C].Proceedings of the Thirteenth Conference on Computational Natural Language Learning.Stroudsburg,PA:ACL,2009:147-155.
[16] Zhao W,Ye J,Yang M,et al.Investigating capsule networks with dynamic routing for text classification[C].Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing.Stroudsburg,PA:Association for Computational Linguistics,2018:3110-3119.
[17] Gong J,Qiu X,Wang S,et al.Information aggregation via dynamic routing for sequence encoding[C].Proceedings of the 27th International Conference on Computational Linguistics.Stroudsburg,PA:Association for Computational Linguistics,2018:2742-2752.
[18] John Lafferty,Andrew McCallum,Fernando Pereira.Conditional random fields:probabilistic models for segmenting and labeling sequence data[C].Proceedings of the 8th International Conference of Machine Learning.San Mateo,CA:Morgan Kaufmann,2001:282-289.
[19] Wang W,Pan S J,Dahlmeier D,et al.Coupled multi-layer attentions for co-extraction of aspect and opinion terms[C].Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence.Menlo Park,CA:AAAI Press,2017:3316-3322.
[20] Zhang X,Jiang Y,Peng H,et al.Semi-supervised structured prediction with neural CRF autoencoder[C].Proceedings of the 2017 conference on empirical methods in natural language processing(EMNLP 2017).Stroudsburg,PA:Association for Computational Linguistics,2017:1701-1711.
[21] Li X,Bing L,Li P,et al.Aspect term extraction with history attention and selective transformation[C].Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence(IJCAI 2018).San Mateo,CA:Morgan Kaufmann,2018:4194-4200.
[22] Xu H,Liu B,Shu L,et al.Double embeddings and cnn-based sequence labeling for aspect extraction[C].Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics.Stroudsburg,PA:Association for Computational Linguistics,2018:592-598.
备注/Memo
收稿日期:2020-05-19