Discourse-Aware Emotion Cause Extraction in Conversations Dexin Kong1 Nan Yu1 Yun Yuan1 Guohong Fu12 Chen Gong12 1School of Computer Science and Technology Soochow University China

2025-05-03 2 0 518.13KB 10 页 10玖币

侵权投诉

Discourse-Aware Emotion Cause Extraction in Conversations

Dexin Kong1, Nan Yu1, Yun Yuan1, Guohong Fu1,2 ∗

, Chen Gong1,2

1School of Computer Science and Technology, Soochow University, China

2Institute of Artiﬁcial Intelligence, Soochow University, China

sakurakdx@gmail.com;

{nyu,yyuanwind}@stu.suda.edu.cn;

{ghfu, gongchen18}@suda.edu.cn

Abstract

Emotion Cause Extraction in Conversa-

tions (ECEC) aims to extract the utterances

which contain the emotional cause in con-

versations. Most prior research focuses

on modelling conversational contexts with

sequential encoding, ignoring the informa-

tive interactions between utterances and

conversational-speciﬁc features for ECEC.

In this paper, we investigate the importance

of discourse structures in handling utterance

interactions and conversation-speciﬁc fea-

tures for ECEC. To this end, we propose

a discourse-aware model (DAM) for this

task. Concretely, we jointly model ECEC

with discourse parsing using a multi-task

learning (MTL) framework and explicitly

encode discourse structures via gated graph

neural network (gated GNN), integrating rich

utterance interaction information to our model.

In addition, we use gated GNN to further

enhance our ECEC model with conversation-

speciﬁc features. Results on the benchmark

corpus show that DAM outperform the state-

of-the-art (SOTA) systems in the literature.

This suggests that the discourse structure may

contain a potential link between emotional

utterances and their corresponding cause

expressions. It also veriﬁes the effectiveness

of conversational-speciﬁc features. The codes

of this paper will be available on GitHub1.

1 Introduction

Emotion cause extraction in conversations (ECEC)

is an important task in conversation analysis. It has

received increasing attention (Poria et al.,2021;

Wang et al.,2021b) with the open conversational

data deluge on social media platforms. Similar

to text-level emotion cause extraction (ECE) (Lee

et al.,2010), this task aims to identify utterances

that contain explanations for emotional causes in

the given conversation. The right part of Fig. 1

∗

Corresponding author.

1http://github.com/

PA:[Mr. Lee, I was wondering if you and your wife had

plans for next Friday night?]

PB: [Yes, I think we might be free. What did you have in

mind?]

PA: [I’ve mailed a formal written invitation to your home

inviting you to a dinner party at my home.]

PB:That sounds great. Is it casual?

PA:Actually, we thought it would be fun to have a formal

sit-down dinner for the holidays.

PB:That sounds good. It will give my wife a chance to

dress up

PA:Yes, we’ve invited four other couples, so it should be

a good mix of people.

PB:That sounds wonderful. Will there be anyone there

that I know?

PA:Oh yes, everyone we are inviting is from our com-

pany.

Q-ELab QAP Q-Elab QAP Ack

Result CQ

QAP

hybird

happiness

Figure 1: An example of ECEC. Left arcs are predicted

discourse structures. Right arcs are emotion-cause pair

annotations.

shows an example of ECEC. The emotion of the

target utterance “That sounds wonderful. Will there

be anyone there that I know?” is “happiness”. The

cause utterance spans of it are “inviting you to a

dinner party”, “it would be fun”, and “It will give

my wife a chance to dress up”, It suggests that

explicit emotion causes aforementioned could be

helpful for many downstream tasks, such as opinion

mining (Choi et al.,2005;Das and Bandyopadhyay,

2010).

ECE has been investigated intensively since

early research (Lee et al.,2010;Chen et al.,2010).

It can be treated as a classiﬁcation task which re-

quires the contexts of a document as inputs and

determine whether each clause is a cause. Re-

cently, several neural models for ECE have been

proposed (Ding and Kejriwal,2020;Li et al.,

2021c;Hu et al.,2021a), using pre-trained lan-

guage models such as BERT (Devlin et al.,2018)

and RoBERTa (Liu et al.,2019) to represent the

arXiv:2210.14419v1 [cs.CL] 26 Oct 2022

plain texts. Compared with text-level ECE, the

research of emotion cause extraction in conversa-

tion is at its preliminary stage. Poria et al. (2021)

propose this task for the ﬁrst time, and apply neu-

ral approaches of text-level emotion cause extrac-

tion (Wei et al.,2020;Ding et al.,2020a,b).

Above traditional approaches treat conversa-

tional contexts as common plain text while us-

ing a Transformer-based encoder (Liu et al.,2019)

to learn the conversational context representation

for the ECEC task (Poria et al.,2021). These

approaches ignore the interaction information be-

tween utterances and conversational-speciﬁc fea-

tures. The problem could be serious when perform-

ing the ECEC task in long conversations.

The discourse structures in a conversation rep-

resent the interactive relationships between utter-

ances. Intuitively, these structures contain the po-

tential links between emotional utterances and their

corresponding cause expressions. As shown by the

solid arc on the left in Fig. 1, most of the emotion-

cause pairs are linked with the target utterance by

discourse relations. In addition, several ﬁndings

of previous studies on ECE support our point of

view. For instance, Hu et al. (2021b); Ding et al.

(2019) believe that the interaction features between

sentences could be useful information for the ECE

task.

In this paper, we propose a discourse-aware

model for ECEC. Concretely, we model ECEC and

discourse parsing in conversations (Afantenos et al.,

2015) jointly. It uses a shared pre-trained language

model (PLM) to represent conversational contexts.

The discourse parsing task can integrate rich ut-

terance interactions information into the shared

PLM. Besides, we use a gated graph neural net-

work (gated GNN) (Li et al.,2016) to explicitly

encode discourse structures that generated by the

discourse parser. In addition, we follow Wang et al.

(2021a), exploiting a gated GNN to further inte-

grate conversation-speciﬁc features such as the rel-

ative utterance distance and speakers to our model.

We conduct the experiments on the standard

benchmark dataset of the ECEC task to verify our

DAM model. Experiments show that the utterance

interaction features are effective for this task. When

the conversation-speciﬁc features are integrated by

gated GNN, the proposed model is able to obtain

further improvements.

To sum up, we make three main contributions as

follows:

•

We propose a discourse-aware model for

ECEC named DAM using multiple task learn-

ing and a gated GNN, which is able to inte-

grate rich utterance interaction features for

ECEC.

•

We further exploit a gated GNN to cap-

ture conversation-speciﬁc features such as the

relative utterance distance and speakers for

ECEC.

•

We advance the performance of the SOTA

models for ECEC.

We organize the rest of this paper as follows.

First, in Section 2, we introduce the related work.

Following in Section 3, we introduce the proposed

model, including the multi-task learning frame-

work, and the gated GNN module. Section 4and

Section 5describes our experiments on a bench-

mark dataset, verifying the effectiveness of our

proposed approach. Finally, we make conclusions

on Section 6.

2 Related Work

Prior research on emotion cause extraction can

be divided into two categories according to the

source text type: text-level emotion cause extrac-

tion (ECE) and emotion cause extraction in conver-

sation (ECEC) (Poria et al.,2021). For the ECE

task, early research adopts linguistic rules (Lee

et al.,2010;Chen et al.,2010;Russo et al.,2011)

and traditional machine learning (Gui et al.,2014,

2016,2018;Xu et al.,2017). In recently years,

several neural network models are introduced to

the ECE task with different granularities, such as

clause-level (Diao et al.,2020;Ding and Kejriwal,

2020;Hu et al.,2021a) and span-level (Li et al.,

2021c,a;Qian et al.,2021;Turcan et al.,2021;

Li et al.,2021b). Except for the text-level emo-

tion cause extraction, Poria et al. (2021) introduces

a new conversational dataset RECCON and the

Transformer-based baseline models.

Many researchers realize that there is a connec-

tion between ECE task and emotion recognition

task. In order to make full use of the interaction

between tasks, researchers use the multi-task frame-

work to carry out ECE task (Chen et al.,2018;Wu

et al.,2020). In addition, researchers take causal

reasoning as an auxiliary task (Fan et al.,2020;

Turcan et al.,2021) to enhance the reasoning abil-

ity of the model. However, most of these methods

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

Discourse-AwareEmotionCauseExtractioninConversationsDexinKong1,NanYu1,YunYuan1,GuohongFu1,2,ChenGong1,21SchoolofComputerScienceandTechnology,SoochowUniversity,China2InstituteofArticialIntelligence,SoochowUniversity,Chinasakurakdx@gmail.com;{nyu,yyuanwind}@stu.suda.edu.cn;{ghfu,gongchen18}@suda.edu...

展开>> 收起<<

Discourse-Aware Emotion Cause Extraction in Conversations Dexin Kong1 Nan Yu1 Yun Yuan1 Guohong Fu12 Chen Gong12 1School of Computer Science and Technology Soochow University China.pdf

共10页,预览2页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Discourse-Aware Emotion Cause Extraction in Conversations Dexin Kong1 Nan Yu1 Yun Yuan1 Guohong Fu12 Chen Gong12 1School of Computer Science and Technology Soochow University China

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: