Document Ground-truth Inter-EC SLSN-U Ms.Huang married a guy 20 years younger than her(c).In order to avoid losing her wealth (c5,c4) None (Cc5,C4) (c2),they have notarized their property before marriage(cs).However the man stole Ms.Huang's money twice after marriage (c),and Ms.Huang submitted a case to the court helplessly(cs). Chen and his wife have two boys and one daughter(c1).The 18-year-old son works in Taizhong (C4,C3) (C4,C3 (c4,c3) (e2).When he knew his mother had been killed by his father(cs),he was quite emotional (c4) (c4,c5), Then he saw his younger brother and sister(cs),and they cried together(c6). (c6.c5) Table 3:Two examples for the case study precision,recall,and F1-score)of SLSN-U exhibit a steady trend when a E[0.1,0.9].This indicates that SLSN-U is robust to the setting of parameter a. From the left subfigure of Figure 5(b),we can find that when B [0.1,0.9],the performance of all three models exhibit a upward trend as B increases.This means that the prediction of E-label and C- label plays a more important role than the prediction of LC-label and LE-label in the training process of SLSN.Again,in most cases,SLSN-U achieves the best performance,and SLSN-E performs better than SLSN-C.From the right subfigure of Figure 5(b),we can find that the recall and F1-score of SLSN-U exhibit an upward trend and the precision of SLSN-U exhibits a downward trend as B increases.This implies that the LPS tends to extract more emotion-cause pairs as B increases. 3.6 Case Study For the case study,we select two examples in the test dataset to demonstrate the effectiveness of our model.The ground-truth and the predicted results of Inter-EC and SLSN-U are given in Table 3. For the first example,Inter-EC predicts None while SLSN-U predicts the correct emotion-cause pair (c5,c4).For the Inter-EC method,it outputs the right pair only when the emotion clause set includes cs and the cause clause set includes c4.While for our method,we can extract emotion-cause pair according to emotion or cause clause.It is easier to establish a matching relationship. For the second example,we can observe that many wrong answers are predicted by Inter-EC (e.g., (c4,c5),(c6,c5)).Due to the use of Cartesian product operation,the connection between emotion clause and cause clause may be ignored,thus many irrelevant pairs will be introduced.It indicates that Cartesian product brings a lot of redundancy and the filter operation fails to filter out the irrelevant pairs.In our method,we can extract emotion-cause pair straightly in the local context window.So our method avoids the redundancy brought by Cartesian product.This is a reason that our method is better than Inter-EC method. 4 Related Work Emotion cause analysis has been studied for about a decade (Lee et al.,2010;Gui et al.,2016;Xia and Ding,2019).Previous studies on emotion cause analysis mainly focused on the emotion cause extraction (ECE)task (Ding et al.,2019;Xia et al.,2019;Fan et al.,2019).Recently,based on the ECE task,a new and more challenging task named emotion cause pair extraction (ECPE)was proposed (Xia and Ding, 2019). The ECE task was first proposed by Lee et al.(2010)and was formalized as a word-level sequence labeling problem.But Chen et al.(2010)suggested that clause may be a more appropriate unit than word for detecting cause.Later,based on this idea,Gui et al.(2016)released a Chinese ECE corpus from a public SINA city news.In this corpus,the ECE task was defined as a clause-level sequence labeling problem,the objective of which is to predict the cause clauses in a document given the emotion.For the following studies on ECE task,this corpus has become a benchmark dataset.While early studies mainly adopted the rule-based methods (Chen et al.,2010;Gao et al.,2015;Gui et al.,2014)and machine learning methods(Ghazi et al.,2015)to deal with the ECE task,recent studies has begun to apply the deep learning methods to this task (Gui et al.,2017;Li et al.,2018;Chen et al.,2018;Ding et al.,2019; Yu et al.,2019;Li et al.,2019;Fan et al.,2019;Xia et al.,2019). Although the ECE task is valuable in practice,its application in real-world scenarios is limited due to the reason that the emotion clauses are naturally not annotated.Considering this situation,Xia and 147147 Document Ground-truth Inter-EC SLSN-U Ms. Huang married a guy 20 years younger than her (c1). In order to avoid losing her wealth (c2), they have notarized their property before marriage (c3). However the man stole Ms. Huang’s money twice after marriage (c4), and Ms. Huang submitted a case to the court helplessly (c5). (c5, c4) None (c5, c4) Chen and his wife have two boys and one daughter (c1). The 18-year-old son works in Taizhong (c2). When he knew his mother had been killed by his father (c3), he was quite emotional (c4). Then he saw his younger brother and sister (c5), and they cried together (c6). (c4, c3) (c4, c3), (c4, c5), (c6, c5) (c4, c3) Table 3: Two examples for the case study precision, recall, and F1-score) of SLSN-U exhibit a steady trend when α ∈ [0.1, 0.9]. This indicates that SLSN-U is robust to the setting of parameter α. From the left subfigure of Figure 5(b), we can find that when β ∈ [0.1, 0.9], the performance of all three models exhibit a upward trend as β increases. This means that the prediction of E-label and C￾label plays a more important role than the prediction of LC-label and LE-label in the training process of SLSN. Again, in most cases, SLSN-U achieves the best performance, and SLSN-E performs better than SLSN-C. From the right subfigure of Figure 5(b), we can find that the recall and F1-score of SLSN-U exhibit an upward trend and the precision of SLSN-U exhibits a downward trend as β increases. This implies that the LPS tends to extract more emotion-cause pairs as β increases. 3.6 Case Study For the case study, we select two examples in the test dataset to demonstrate the effectiveness of our model. The ground-truth and the predicted results of Inter-EC and SLSN-U are given in Table 3. For the first example, Inter-EC predicts None while SLSN-U predicts the correct emotion-cause pair (c5, c4). For the Inter-EC method, it outputs the right pair only when the emotion clause set includes c5 and the cause clause set includes c4. While for our method, we can extract emotion-cause pair according to emotion or cause clause. It is easier to establish a matching relationship. For the second example, we can observe that many wrong answers are predicted by Inter-EC (e.g., (c4, c5), (c6, c5)). Due to the use of Cartesian product operation, the connection between emotion clause and cause clause may be ignored, thus many irrelevant pairs will be introduced. It indicates that Cartesian product brings a lot of redundancy and the filter operation fails to filter out the irrelevant pairs. In our method, we can extract emotion-cause pair straightly in the local context window. So our method avoids the redundancy brought by Cartesian product. This is a reason that our method is better than Inter-EC method. 4 Related Work Emotion cause analysis has been studied for about a decade (Lee et al., 2010; Gui et al., 2016; Xia and Ding, 2019). Previous studies on emotion cause analysis mainly focused on the emotion cause extraction (ECE) task (Ding et al., 2019; Xia et al., 2019; Fan et al., 2019). Recently, based on the ECE task, a new and more challenging task named emotion cause pair extraction (ECPE) was proposed (Xia and Ding, 2019). The ECE task was first proposed by Lee et al. (2010) and was formalized as a word-level sequence labeling problem. But Chen et al. (2010) suggested that clause may be a more appropriate unit than word for detecting cause. Later, based on this idea, Gui et al. (2016) released a Chinese ECE corpus from a public SINA city news. In this corpus, the ECE task was defined as a clause-level sequence labeling problem, the objective of which is to predict the cause clauses in a document given the emotion. For the following studies on ECE task, this corpus has become a benchmark dataset. While early studies mainly adopted the rule-based methods (Chen et al., 2010; Gao et al., 2015; Gui et al., 2014) and machine learning methods (Ghazi et al., 2015) to deal with the ECE task, recent studies has begun to apply the deep learning methods to this task (Gui et al., 2017; Li et al., 2018; Chen et al., 2018; Ding et al., 2019; Yu et al., 2019; Li et al., 2019; Fan et al., 2019; Xia et al., 2019). Although the ECE task is valuable in practice, its application in real-world scenarios is limited due to the reason that the emotion clauses are naturally not annotated. Considering this situation, Xia and