Pairwise tagging framework for end-to-end emotion-cause pair extraction

doi:10.1007/s11704-022-1409-x

Front. Comput. Sci.

2023, Vol. 17

Issue (2) : 172314 https://doi.org/10.1007/s11704-022-1409-x

RESEARCH ARTICLE

Pairwise tagging framework for end-to-end emotion-cause pair extraction

Zhen WU^1,², Xinyu DAI^1,²(

), Rui XIA³

¹. National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210023, China
². Collaborative Innovation Center of Novel Software Technology and Industrialization, Nanjing 210023, China
³. School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing 210023, China

Download: PDF(2697 KB) HTML
Export: BibTeX | EndNote | Reference Manager | ProCite | RefWorks

Abstract

Emotion-cause pair extraction (ECPE) aims to extract all the pairs of emotions and corresponding causes in a document. It generally contains three subtasks, emotions extraction, causes extraction, and causal relations detection between emotions and causes. Existing works adopt pipelined approaches or multi-task learning to address the ECPE task. However, the pipelined approaches easily suffer from error propagation in real-world scenarios. Typical multi-task learning cannot optimize all tasks globally and may lead to suboptimal extraction results. To address these issues, we propose a novel framework, Pairwise Tagging Framework (PTF), tackling the complete emotion-cause pair extraction in one unified tagging task. Unlike prior works, PTF innovatively transforms all subtasks of ECPE, i.e., emotions extraction, causes extraction, and causal relations detection between emotions and causes, into one unified clause-pair tagging task. Through this unified tagging task, we can optimize the ECPE task globally and extract more accurate emotion-cause pairs. To validate the feasibility and effectiveness of PTF, we design an end-to-end PTF-based neural network and conduct experiments on the ECPE benchmark dataset. The experimental results show that our method outperforms pipelined approaches significantly and typical multi-task learning approaches.

Keywords emotion-cause pair extraction pairwise tagging framework end-to-end neural network

Corresponding Author(s): Xinyu DAI

Just Accepted Date: 02 December 2021 Issue Date: 02 August 2022

Cite this article:

Zhen WU,Xinyu DAI,Rui XIA. Pairwise tagging framework for end-to-end emotion-cause pair extraction[J]. Front. Comput. Sci., 2023, 17(2): 172314.

URL:

https://academic.hep.com.cn/fcs/EN/10.1007/s11704-022-1409-x
https://academic.hep.com.cn/fcs/EN/Y2023/V17/I2/172314

Fig.1 An example of the ECPE task. There are six clauses in the document. The clauses highlighted in bold are emotion clauses, and the clauses in underline are cause clauses. The goal of ECPE is to extract three emotion-cause pairs shown in the lower part

Tags	Meanings
$E$	the clause-pair $(c i, c i)$ expresses emotion, i.e., $c i$ is an emotion clause.
$C$	the clause-pair $(c j, c j)$ expresses cause, i.e., $c j$ is a cause clause.
$P$	the clause-pair $(c i, c j)$ contains a casual relation.
$O$	no above three relations for clause-pair $(c i, c j) .$

Tab.1 The meanings of PTF tags for the ECPE task

Fig.2 PTF tagging result of example document of Fig.1. There are three emotion-cause pairs

(c 4, c 2)

(c 4, c 3)

, and

(c 5, c 6)

in the example document

Fig.3 An illustration of pairwise tagging network (PTN)

Tab.2 Statistics of the ECPE dataset

Hyper-parameters	Values
Dimension of word embedding	200
Dimension of LSTM cell	100
Dimension of position embedding	50
Dropout rate	0.5
Batch size	32
Learning rate	0.005
Maximum words of a clause	75
Maximum clauses of a document	45
Loss weight $λ e$	1
Loss weight $λ c$	1

Tab.3 Hyper-parameter settings

Tab.4 Main experiment results of different methods (%). Best and second-best results are respectively in bold and underline

Tab.5 Experiment results of different methods using BERT as the encoder (%). Best and second-best results are respectively in bold and underline

Tab.6 Ablation study of removing CPE, ECP or CSA respectively from the complete model PTN (%). Best results are in bold

Document	Ground truth	RankCP	PTN
$c 1$ :[ $T h e h e a d m a s t e r p e r s i s t e d i n s t a n d i n g o n t h e p o d i u m w i t h h i s s t u m p e d l e g s f o r 34 y e a r s,_$ ] $c 2$ :[ $a n d c a r e d_$ $a b o u t h i s s t u d e n t s w h e n h e w a s c r i t i c a l l y i l l ._$ ] $c 3$ :[The villagers expressed their highest respect for him.]	$(c 3, c 1)$ $(c 3, c 2)$	$(c 3, c 2)$ √	$(c 3, c 1)$ √ $(c 3, c 2)$ √
$c 1$ :[A group of caring volunteers surrounded the couple of Li Shiming and their two-year-old son Li Muyixin and walked into the house happily.] $c 2$ :[For Li Shiming’s family], $c 3$ :[that day was a happy day], $c 4$ :[ $a n d t h e_$ $s o n w h o h a d b e e n a b d u c t e d f o r 45 d a y s f i n a l l y r e t u r n e d h o m e t h a t d a y ._$ ]	$(c 3, c 4)$	$(c 3, c 4)$ √ $(c 1, c 3)$ ×	$(c 3, c 4)$ √

Tab.7 Examples of predicted emotion-cause pairs of RankCP and PTN. The clauses in bold and underline respectively denote the emotions and causes

Fig.4 The performance of PTN on the ECPE task with different document lengths. The proportions of document length ranges (0, 10], (10, 20], (30, 40], (30, +) are respectively 21.53%, 62.56%, 13.33%, and 2.58% in the testing set

1	E Fox. Emotion Science: Cognitive and Neuroscientific Approaches to Understanding Human Emotions. New York: Palgrave Macmillan, 2008
2	T, Brosch K R, Scherer D, Grandjean D Sander . The impact of emotion on perception, attention, memory, and decision-making. Swiss Medical Weekly, 2013, 143: w13786
3	C, Quan F Ren. Construction of a blog emotion corpus for Chinese emotional expression analysis. In: Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing. 2009, 1446– 1454
4	J R Bellegarda. Emotion analysis using latent affective folding and embedding. In: Proceedings of the 2010 NAACL HLT Workshop on Computational Approaches to Analysis and Generation of Emotion in Text. 2010, 1– 9
5	A, Qadir E Riloff. Learning emotion indicators from tweets: hashtags, hashtag patterns, and phrases. In: Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing. 2014, 1203– 1209
6	S Y M, Lee Y, Chen C R Huang. A text-driven rule-based system for emotion cause detection. In: Proceedings of the 2010 NAACL HLT Workshop on Computational Approaches to Analysis and Generation of Emotion in Text. 2010, 45– 53
7	I, Russo T, Caselli F, Rubino E, Boldrini P Martínez-Barco. EMOCause: an easy-adaptable approach to extract emotion cause contexts. In: Proceedings of the 2nd Workshop on Computational Approaches to Subjectivity and Sentiment Analysis. 2011, 153– 160
8	A, Neviarouskaya M Aono. Extracting causes of emotions from text. In: Proceedings of the 6th International Joint Conference on Natural Language Processing. 2013, 932– 936
9	K, Gao H, Xu J Wang . A rule-based approach to emotion cause detection for Chinese micro-blogs. Expert Systems with Applications, 2015, 42( 9): 4517– 4528
10	L, Gui D, Wu R, Xu Q, Lu Y Zhou. Event-driven emotion cause extraction with corpus construction. In: Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing. 2016, 1639– 1649
11	X, Cheng Y, Chen B, Cheng S, Li G Zhou . An emotion cause corpus for Chinese microblogs with multiple-user structures. ACM Transactions on Asian and Low-Resource Language Information Processing, 2018, 17( 1): 6
12	C, Fan H, Yan J, Du L, Gui L, Bing M, Yang R, Xu R Mao. A knowledge regularized hierarchical approach for emotion cause analysis. In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing. 2019, 5614– 5624
13	R, Xia Z Ding. Emotion-cause pair extraction: a new task to emotion analysis in texts. In: Proceedings of the 57th Conference of the Association for Computational Linguistics. 2019, 1003– 1012
14	H, Song C, Zhang Q, Li D Song. End-to-end emotion-cause pair extraction via learning to link. 2020, arXiv preprint arXiv: 2002.10710
15	H, Tang D, Ji Q Zhou . Joint multi-level attentional model for emotion detection and emotion-cause pair extraction. Neurocomputing, 2020, 409: 329– 340
16	Z, Ding R, Xia J Yu. ECPE-2D: emotion-cause pair extraction based on joint two-dimensional representation, interaction and prediction. In: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020, 3161– 3170
17	P, Wei J, Zhao W Mao. Effective inter-clause modeling for end-to-end emotion-cause pair extraction. In: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020, 3171– 3181
18	S Y M, Lee Y, Chen C R, Huang S S Li . Detecting emotion causes with a linguistic rule-based approach. Computational Intelligence, 2013, 29( 3): 390– 416
19	W, Li H Xu . Text-based emotion classification using emotion cause extraction. Expert Systems with Applications, 2014, 41( 4): 1742– 1749
20	K, Gao H, Xu J Wang. Emotion cause detection for Chinese micro-blogs based on ECOCC model. In: Proceedings of the 19th Pacific-Asia Conference on Knowledge Discovery and Data Mining. 2015, 3– 14
21	S, Yada K, Ikeda K, Hoashi K Kageura. A bootstrap method for automatic rule acquisition on emotion cause extraction. In: Proceedings of the 2017 IEEE International Conference on Data Mining Workshops. 2017, 414– 421
22	Y, Chen S Y M, Lee S, Li C R Huang. Emotion cause detection with linguistic constructions. In: Proceedings of the 23rd International Conference on Computational Linguistics. 2010, 179– 187
23	L, Gui L, Yuan R, Xu B, Liu Q, Lu Y Zhou. Emotion cause detection with linguistic construction in Chinese Weibo text. In: Proceedings of the 3rd CCF International Conference on Natural Language Processing and Chinese Computing. 2014, 457– 464
24	D, Ghazi D, Inkpen S Szpakowicz. Detecting emotion stimuli in emotion-bearing sentences. In: Proceedings of the 16th International Conference on Computational Linguistics and Intelligent Text Processing. 2015, 152– 165
25	L, Gui R, Xu Q, Lu D, Wu Y Zhou. Emotion cause extraction, a challenging task with corpus construction. In: Proceedings of the 5th National Conference on Social Media Processing. 2016, 98– 109
26	R, Xu J, Hu Q, Lu D, Wu L Gui . An ensemble approach for emotion cause detection with event extraction and multi-kernel SVMs. Tsinghua Science and Technology, 2017, 22( 6): 646– 659
27	L, Gui J, Hu Y, He R, Xu Q, Lu J Du. A question answering approach to emotion cause extraction. 2017, arXiv preprint arXiv: 1708.05482
28	X, Li K, Song S, Feng D, Wang Y Zhang. A co-attention neural network model for emotion cause analysis with emotional context awareness. In: Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. 2018, 4752– 4757
29	X, Yu W, Rong Z, Zhang Y, Ouyang Z Xiong . Multiple level hierarchical network-based clause selection for emotion cause extraction. IEEE Access, 2019, 7: 9071– 9079
30	Z, Ding H, He M, Zhang R Xia. From independent prediction to reordered prediction: integrating relative position and global label information to emotion cause identification. In: Proceedings of the 33rd AAAI Conference on Artificial Intelligence. 2019, 6343– 6350
31	R, Xia M, Zhang Z Ding. RTHN: a RNN-transformer hierarchical network for emotion cause extraction. In: Proceedings of the 28th International Joint Conference on Artificial Intelligence. 2019, 5285– 5291
32	S, Wu F, Chen F, Wu Y, Huang X Li. A multi-task learning neural network for emotion-cause pair extraction. In: Proceedings of the 24th European Conference on Artificial Intelligence. 2020, 2212– 2219
33	C, Fan C, Yuan L, Gui Y, Zhang R Xu . Multi-task sequence tagging for emotion-cause pair extraction via tag distribution refinement. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2021, 29: 2339– 2350
34	Z, Cheng Z, Jiang Y, Yin H, Yu Q Gu. A symmetric local search network for emotion-cause pair extraction. In: Proceedings of the 28th International Conference on Computational Linguistics. 2020, 139– 149
35	J, Yu W, Liu Y, He C Zhang . A mutually auxiliary multitask model with self-distillation for emotion-cause pair extraction. IEEE Access, 2021, 9: 26811– 26821
36	C, Fan C, Yuan J, Du L, Gui M, Yang R Xu. Transition-based directed graph construction for emotion-cause pair extraction. In: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020, 3707– 3717
37	C, Yuan C, Fan J, Bao R Xu. Emotion-cause pair extraction as sequence labeling based on a novel tagging scheme. In: Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing. 2020, 3568– 3573
38	Z, Cheng Z, Jiang Y, Yin N, Li Q Gu . A unified target-oriented sequence-to-sequence model for emotion-cause pair extraction. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2021, 29: 2779– 2791
39	Z, Yang D, Yang C, Dyer X, He A, Smola E Hovy. Hierarchical attention networks for document classification. In: Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2016, 1480− 1489
40	Z, Wu X Y, Dai C, Yin S, Huang J Chen. Improving review representations with user attention and product attention for sentiment classification. In: Proceedings of the 32nd AAAI Conference on Artificial Intelligence. 2018, 5989− 5996
41	S, Hochreiter J Schmidhuber . Long short-term memory. Neural Computation, 1997, 9( 8): 1735– 1780
42	D, Bahdanau K, Cho Y Bengio. Neural machine translation by jointly learning to align and translate. In: Proceedings of the 3rd International Conference on Learning Representations. 2015
43	J, Devlin M W, Chang K, Lee K Toutanova. BERT: pre-training of deep bidirectional transformers for language understanding. In: Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2019, 4171− 4186
44	T, Mikolov I, Sutskever K, Chen G, Corrado J Dean. Distributed representations of words and phrases and their compositionality. In: Proceedings of the 26th International Conference on Neural Information Processing Systems. 2013, 3111− 3119
45	N, Srivastava G, Hinton A, Krizhevsky I, Sutskever R Salakhutdinov . Dropout: a simple way to prevent neural networks from overfitting. The Journal of Machine Learning Research, 2014, 15( 1): 1929– 1958
46	D P, Kingma J Ba. Adam: a method for stochastic optimization. In: Proceedings of the 3rd International Conference on Learning Representations. 2015
47	A, Vaswani N, Shazeer N, Parmar J, Uszkoreit L, Jones A N, Gomez Ł, Kaiser I Polosukhin. Attention is all you need. In: Proceedings of the 31st International Conference on Neural Information Processing Systems. 2017, 6000− 6010

[1]

FCS-21409-OF-ZW_suppl_1

Download

[1]	Miao ZHANG, Tingting HE, Ming DONG. Meta-path reasoning of knowledge graph for commonsense question answering[J]. Front. Comput. Sci., 2024, 18(1): 181303-.
[2]	Shuo TAN, Lei ZHANG, Xin SHU, Zizhou WANG. A feature-wise attention module based on the difference with surrounding features for convolutional neural networks[J]. Front. Comput. Sci., 2023, 17(6): 176338-.
[3]	Yongquan LIANG, Qiuyu SONG, Zhongying ZHAO, Hui ZHOU, Maoguo GONG. BA-GNN: Behavior-aware graph neural network for session-based recommendation[J]. Front. Comput. Sci., 2023, 17(6): 176613-.
[4]	Yamin HU, Hao JIANG, Zongyao HU. Measuring code maintainability with deep neural networks[J]. Front. Comput. Sci., 2023, 17(6): 176214-.
[5]	Jinwei LUO, Mingkai HE, Weike PAN, Zhong MING. BGNN: Behavior-aware graph neural network for heterogeneous session-based recommendation[J]. Front. Comput. Sci., 2023, 17(5): 175336-.
[6]	Yuan GAO, Xiang WANG, Xiangnan HE, Huamin FENG, Yongdong ZHANG. Rumor detection with self-supervised learning on texts and social graph[J]. Front. Comput. Sci., 2023, 17(4): 174611-.
[7]	Shuang LIU, Fan ZHANG, Baiyang ZHAO, Renjie GUO, Tao CHEN, Meishan ZHANG. APPCorp: a corpus for Android privacy policy document structure analysis[J]. Front. Comput. Sci., 2023, 17(3): 173320-.
[8]	Zhe XUE, Junping DU, Xin XU, Xiangbin LIU, Junfu WANG, Feifei KOU. Few-shot node classification via local adaptive discriminant structure learning[J]. Front. Comput. Sci., 2023, 17(2): 172316-.
[9]	Hongjia RUAN, Huihui SONG, Bo LIU, Yong CHENG, Qingshan LIU. Intellectual property protection for deep semantic segmentation models[J]. Front. Comput. Sci., 2023, 17(1): 171306-.
[10]	Tian WANG, Jiakun LI, Huai-Ning WU, Ce LI, Hichem SNOUSSI, Yang WU. ResLNet: deep residual LSTM network with longer input for action recognition[J]. Front. Comput. Sci., 2022, 16(6): 166334-.
[11]	Yunyun WANG, Chao WANG, Hui XUE, Songcan CHEN. Self-corrected unsupervised domain adaptation[J]. Front. Comput. Sci., 2022, 16(5): 165323-.
[12]	Guoshuai ZHOU, Xiuxia TIAN, Aoying ZHOU. Image copy-move forgery passive detection based on improved PCNN and self-selected sub-images[J]. Front. Comput. Sci., 2022, 16(4): 164705-.
[13]	Ashok KUMAR, Arpit JAIN. Image smog restoration using oblique gradient profile prior and energy minimization[J]. Front. Comput. Sci., 2021, 15(6): 156706-.
[14]	Huiying ZHANG, Yu ZHANG, Xin GENG. Practical age estimation using deep label distribution learning[J]. Front. Comput. Sci., 2021, 15(3): 153318-.
[15]	Yixuan CAO, Dian CHEN, Zhengqi XU, Hongwei LI, Ping LUO. Nested relation extraction with iterative neural network[J]. Front. Comput. Sci., 2021, 15(3): 153323-.

Viewed

Full text

Abstract

Cited

Shared

Discussed