arxiv: 2604.19547 · v1 · submitted 2026-04-21 · 💻 cs.CL

Recognition: unknown

Emotion-Cause Pair Extraction in Conversations via Semantic Decoupling and Graph Alignment

Tianxiang Ma , Weijie Feng , Xinyu Wang , Zhiyong Cheng

Authors on Pith no claims yet

Pith reviewed 2026-05-10 01:57 UTC · model grok-4.3

classification 💻 cs.CL

keywords emotion-cause pair extractionsemantic decouplingoptimal transportgraph alignmentconversational causalitydialogue understandingmany-to-many relations

0 comments

The pith

Disentangling emotion semantics from cause semantics and aligning them globally via optimal transport extracts consistent many-to-many emotion-cause pairs in conversations.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper seeks to improve Emotion-Cause Pair Extraction in Conversations by rejecting independent pairwise classification. It separates emotion-oriented semantics from cause-oriented semantics into two complementary representation spaces so each can play its distinct role in how feelings spread and explanations are given. These spaces are then aligned through optimal transport to produce globally consistent matches that handle cases with multiple causes for one emotion or vice versa. Prior methods missed this because they checked pairs locally without regard for the full dialogue structure. The SCALE framework implements the decoupling and alignment together and records stronger results than earlier approaches on standard test sets.

Core claim

Reformulating ECPEC as a global alignment problem between decoupled emotion-side and cause-side representations, solved via optimal transport inside a shared conversational structure, enables many-to-many and globally consistent emotion-cause matching.

What carries the argument

Semantic decoupling of emotion and cause representations into complementary spaces followed by optimal transport alignment within the SCALE framework.

If this is right

Many-to-many causal relations become directly modelable without forcing one-to-one assumptions.
Global consistency across an entire dialogue replaces local pairwise decisions.
State-of-the-art results appear on multiple ECPEC benchmark datasets.
The same decoupling-plus-alignment pattern can be inserted into other shared conversational encoders.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The dual-space idea may transfer to other dialogue tasks that separate affective content from explanatory content, such as intent or opinion triggering.
If the two spaces remain complementary, similar splits could be tried for dual-aspect problems outside emotion, such as separating topic from sentiment.
Longer multi-party dialogues are a natural next test bed where global alignment would be expected to show larger gains than on short exchanges.

Load-bearing premise

Emotion-oriented and cause-oriented semantics can be disentangled into two complementary representation spaces that capture their distinct conversational roles without losing essential shared information.

What would settle it

A controlled run on the same benchmark datasets in which removing the semantic decoupling step and using a single shared representation yields equal or higher accuracy on many-to-many pairs would show the separation is not required.

Figures

Figures reproduced from arXiv: 2604.19547 by Tianxiang Ma, Weijie Feng, Xinyu Wang, Zhiyong Cheng.

**Figure 1.** Figure 1: Comparison between the existing ECPEC paradigm (b-c) and SCALE (d). recognition in conversation (ERC) (Wang et al., 2024b; Fu et al., 2023; Majumder et al., 2019), which focuses on assigning discrete emotion labels to individual utterances (Fu et al., 2021; Hu et al., 2021), ECPEC provides a causal perspective for dialogue understanding, enabling more fine-grained analysis of emotional dynamics. As highli… view at source ↗

**Figure 2.** Figure 2: Overall architecture of our SCALE. An overview of the proposed SCALE is illustrated in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Figure 3: Qualitative analysis. fies the dual-cause emotion [(u4, u3),(u4, u4)], but fails to capture both causes in [(u6, u6),(u6, u3)]. These observations suggest that while the global alignment mechanism enables flexible one-to-many reasoning, modeling long-range causal dependencies remains an open challenge. Error Analysis. To further analyze the behavior of SCALE in multi-cause scenarios, we examine its perfor… view at source ↗

**Figure 4.** Figure 4: t-SNE visualization of emotion and cause [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗

**Figure 5.** Figure 5: The system prompt template. D Computational Complexity Analysis Theoretically, for a conversation with N utterances and |E| edges, one GNN layer costs O(|E|d). With L layers and two encoders, encoding cost is O(2L|E|d), i.e., only a constant-factor increase. The Sinkhorn alignment operates on an N × N matrix with per-iteration cost O(N2 ). Since dialogue length is small (N < 30), the alignment remains li… view at source ↗

read the original abstract

Emotion-Cause Pair Extraction in Conversations (ECPEC) aims to identify the set of causal relations between emotion utterances and their triggering causes within a dialogue. Most existing approaches formulate ECPEC as an independent pairwise classification task, overlooking the distinct semantics of emotion diffusion and cause explanation, and failing to capture globally consistent many-to-many conversational causality. To address these limitations, we revisit ECPEC from a semantic perspective and seek to disentangle emotion-oriented semantics from cause-oriented semantics, mapping them into two complementary representation spaces to better capture their distinct conversational roles. Building on this semantic decoupling, we naturally formulate ECPEC as a global alignment problem between the emotion-side and cause-side representations, and employ optimal transport to enable many-to-many and globally consistent emotion-cause matching. Based on this perspective, we propose a unified framework SCALE that instantiates the above semantic decoupling and alignment principle within a shared conversational structure. Extensive experiments on several benchmark datasets demonstrate that SCALE consistently achieves state-of-the-art performance. Our codes are released at https://github.com/CoCoSphere/SCALE.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

SCALE reframes ECPEC with semantic decoupling plus optimal transport alignment and hits SOTA, but the decoupling step lacks clear enforcement or proof that the spaces stay distinct.

read the letter

The paper's core move is to stop treating emotion-cause pairs as independent classifications and instead split the semantics into two spaces inside a shared graph, then align them globally with optimal transport for many-to-many matches. That formulation is new for this task and directly targets the conversational structure that prior work ignored. The experiments claim consistent gains on the usual benchmarks, and the code release is a plus for anyone who wants to inspect or extend it. The graph component and the alignment step look like the parts doing real work, and the authors are straightforward about building on existing conversational modeling rather than claiming a total reinvention. The main weakness is that nothing in the setup forces the two spaces to remain complementary. Without an orthogonality term, contrastive loss, or similar check, the emotion and cause branches could simply learn overlapping features, and any improvement might trace back to the graph or extra context rather than the decoupling itself. The abstract is thin on ablation details and exact baselines, so the full paper needs to show that removing the split or the transport step actually hurts. If those controls are present and the numbers hold, the argument strengthens; if not, the central claim stays under-supported. This is a solid incremental paper for the ECPEC niche. Readers working on dialogue understanding or graph alignment in NLP will find the modeling choice useful to test. It is coherent on its own terms and has enough empirical grounding to merit peer review, though the decoupling premise will need tighter validation in revision.

Referee Report

3 major / 2 minor

Summary. The paper proposes SCALE, a unified framework for Emotion-Cause Pair Extraction in Conversations (ECPEC) that disentangles emotion-oriented semantics from cause-oriented semantics into two complementary representation spaces within a shared conversational graph structure. It formulates ECPEC as a global alignment problem solved via optimal transport to enable many-to-many and consistent emotion-cause matching, and reports state-of-the-art results on benchmark datasets with released code.

Significance. If the semantic decoupling and optimal transport alignment hold, the work could meaningfully advance ECPEC by moving beyond pairwise classification to capture distinct conversational roles and global causality, with the code release aiding reproducibility in conversational NLP.

major comments (3)

[§3] §3 (Method), Semantic Decoupling and Graph Alignment: The central premise requires that emotion and cause semantics are mapped to complementary, non-redundant spaces, yet no auxiliary objective (contrastive loss, orthogonality constraint, or mutual-information term) is described to enforce distinctness between the two branches. Without such regularization, the observed gains could be explained by the graph component or richer context modeling alone, leaving the decoupling claim unverified and load-bearing for the paper's argument.
[§4] §4 (Experiments): The manuscript reports SOTA performance but provides insufficient detail on experimental setup, choice of baselines, hyperparameter tuning, and ablation studies that isolate the contribution of semantic decoupling versus the shared graph or optimal transport. This absence prevents confirmation that the proposed principle, rather than implementation details, drives the improvements.
[Table 2] Table 2 or equivalent results table: While overall F1 scores are highlighted as superior, the lack of per-relation or many-to-many specific metrics (e.g., precision on multi-cause emotions) makes it hard to substantiate the claim of 'globally consistent many-to-many conversational causality' beyond aggregate end-task gains.

minor comments (2)

[Abstract] The abstract and introduction could more explicitly contrast the proposed optimal transport alignment with prior pairwise or sequence-labeling baselines to clarify the novelty.
[§3] Notation for the two representation spaces (e.g., E and C) should be introduced earlier and used consistently to improve readability.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback and the recommendation for major revision. We address each major comment point by point below and will revise the manuscript to improve clarity, rigor, and substantiation of the claims.

read point-by-point responses

Referee: [§3] §3 (Method), Semantic Decoupling and Graph Alignment: The central premise requires that emotion and cause semantics are mapped to complementary, non-redundant spaces, yet no auxiliary objective (contrastive loss, orthogonality constraint, or mutual-information term) is described to enforce distinctness between the two branches. Without such regularization, the observed gains could be explained by the graph component or richer context modeling alone, leaving the decoupling claim unverified and load-bearing for the paper's argument.

Authors: We acknowledge that the submitted manuscript did not describe an explicit auxiliary objective to enforce distinctness between the emotion and cause representation spaces. While the architecture uses separate projection heads and the optimal transport objective encourages complementarity through alignment, this does not directly verify non-redundancy. To address this, we will introduce an orthogonality constraint (or contrastive term) between the two branches in the revised §3 and report the corresponding ablation to isolate its contribution. revision: yes
Referee: [§4] §4 (Experiments): The manuscript reports SOTA performance but provides insufficient detail on experimental setup, choice of baselines, hyperparameter tuning, and ablation studies that isolate the contribution of semantic decoupling versus the shared graph or optimal transport. This absence prevents confirmation that the proposed principle, rather than implementation details, drives the improvements.

Authors: We agree that the experimental section requires more detail to allow verification of the source of gains. In the revision we will expand §4 with complete hyperparameter tables, explicit baseline configurations and sources, and additional ablations that separately disable semantic decoupling, the shared graph, and the optimal transport module while keeping other factors fixed. revision: yes
Referee: [Table 2] Table 2 or equivalent results table: While overall F1 scores are highlighted as superior, the lack of per-relation or many-to-many specific metrics (e.g., precision on multi-cause emotions) makes it hard to substantiate the claim of 'globally consistent many-to-many conversational causality' beyond aggregate end-task gains.

Authors: Overall F1 is the primary reported metric in prior ECPEC work, yet we recognize that aggregate scores alone do not fully demonstrate the many-to-many consistency claim. We will add per-relation F1 scores and a dedicated many-to-many analysis (e.g., precision/recall on multi-cause emotions and multi-emotion causes) to the revised results table and discussion. revision: yes

Circularity Check

0 steps flagged

No circularity: semantic decoupling and alignment presented as architectural design choice with independent empirical validation

full rationale

The paper's derivation begins from a semantic perspective on ECPEC limitations, proposes disentangling emotion-oriented and cause-oriented semantics into complementary spaces, and formulates the task as optimal-transport alignment within a shared graph structure instantiated as SCALE. This chain consists of modeling decisions and a new framework rather than any reduction of outputs to inputs by construction. No equations or claims equate a 'prediction' to a fitted parameter, no uniqueness theorem is imported from self-citations, and no ansatz is smuggled via prior work. The central premise is an explicit design choice whose value is assessed via benchmark experiments, making the argument self-contained against external data.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are detailed beyond standard neural network components and optimal transport.

pith-pipeline@v0.9.0 · 5487 in / 970 out tokens · 24407 ms · 2026-05-10T01:57:20.861299+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

74 extracted references · 48 canonical work pages · 1 internal anchor

[1]

Induced and reduced unbounded operator algebras

CONSK-GCN: Conversational Semantic- and Knowledge-Oriented Graph Convolutional Network for Multimodal Emotion Recognition , shorttitle =. 2021 IEEE International Conference on Multimedia and Expo (ICME) , author =. 2021 , month = jul, pages =. doi:10.1109/ICME51207.2021.9428438 , urldate =

work page Pith review doi:10.1109/icme51207.2021.9428438 2021
[2]

2022 , month = jul, journal =

Context- and Knowledge-Aware Graph Convolutional Network for Multimodal Emotion Recognition , author =. 2022 , month = jul, journal =. doi:10.1109/MMUL.2022.3173430 , urldate =

work page doi:10.1109/mmul.2022.3173430 2022
[3]

2023 , journal =

Emotion Recognition in Conversations: A Survey Focusing on Context, Speaker Dependencies, and Fusion Methods , author =. 2023 , journal =

2023
[4]

INTERNATIONAL CONFERENCE ON MULTIMODAL INTERACTION , author =

GCFormer: A Graph Convolutional Transformer for Speech Emotion Recognition , shorttitle =. INTERNATIONAL CONFERENCE ON MULTIMODAL INTERACTION , author =. 2023 , month = oct, pages =. doi:10.1145/3577190.3614177 , urldate =

work page doi:10.1145/3577190.3614177 2023
[5]

Proceedings of the 14th Workshop on Computational Approaches to Subjectivity, Sentiment, & Social Media Analysis , author =

SEC: Context-Aware Metric Learning for Efficient Emotion Recognition in Conversation , shorttitle =. Proceedings of the 14th Workshop on Computational Approaches to Subjectivity, Sentiment, & Social Media Analysis , author =. 2024 , month = aug, pages =

2024
[6]

DialogueGCN: A Graph Convolutional Neural Network for Emotion Recognition in Conversation , shorttitle =. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) , author =. 2019 , month = nov, pages =. doi:10.18653/v1/D19-1015 , urldate =

work page doi:10.18653/v1/d19-1015 2019
[7]

Benchmarking Cyber Harassment Dialogue Comprehension through Emotion-Informed Manifestations-Determinants Demarcation , booktitle =

Ghosh, Soumitra and Singh, Gopendra Vikram and Arora, Jashn and Ekbal, Asif , year =. Benchmarking Cyber Harassment Dialogue Comprehension through Emotion-Informed Manifestations-Determinants Demarcation , booktitle =. doi:10.1609/aaai.v38i20.30207 , urldate =

work page doi:10.1609/aaai.v38i20.30207
[8]

MMGCN: Multimodal Fusion via Deep Graph Convolution Network for Emotion Recognition in Conversation , shorttitle =. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) , author =. 2021 , month = aug, pages =. doi:10.18653/...

work page doi:10.18653/v1/2021.acl-long.440 2021
[9]

Audioclip: Extending clip to image, text and audio

MM-DFN: Multimodal Dynamic Fusion Network for Emotion Recognition in Conversations , shorttitle =. ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , author =. 2022 , month = may, pages =. doi:10.1109/ICASSP43922.2022.9747397 , urldate =

work page doi:10.1109/icassp43922.2022.9747397 2022
[10]

DailyDialog: A Manually Labelled Multi-Turn Dialogue Dataset , booktitle =

Li, Yanran and Su, Hui and Shen, Xiaoyu and Li, Wenjie and Cao, Ziqiang and Niu, Shuzi , year =. DailyDialog: A Manually Labelled Multi-Turn Dialogue Dataset , booktitle =
[11]

2023 , journal =

GraphMFT: A Graph Network Based Multimodal Fusion Technique for Emotion Recognition in Conversation , author =. 2023 , journal =

2023
[12]

Gelbukh, and Erik Cambria

DialogueRNN: An Attentive RNN for Emotion Detection in Conversations , shorttitle =. Proceedings of the AAAI Conference on Artificial Intelligence , author =. 2019 , month = jul, volume =. doi:10.1609/aaai.v33i01.33016818 , urldate =

work page doi:10.1609/aaai.v33i01.33016818 2019
[13]

Findings of the Association for Computational Linguistics: EMNLP 2021 , author =

DialogueTRM: Exploring Multi-Modal Emotional Dynamics in a Conversation , shorttitle =. Findings of the Association for Computational Linguistics: EMNLP 2021 , author =. 2021 , month = nov, pages =. doi:10.18653/v1/2021.findings-emnlp.229 , urldate =

work page doi:10.18653/v1/2021.findings-emnlp.229 2021
[14]

2024 , month = may, number =

Revisiting Multimodal Emotion Recognition in Conversation from the Perspective of Graph Spectrum , author =. 2024 , month = may, number =. 2404.17862 , publisher =

work page arXiv 2024
[15]

Proceedings of the 32nd ACM International Conference on Multimedia , author =

Ada2I: Enhancing Modality Balance for Multimodal Conversational Emotion Recognition , shorttitle =. Proceedings of the 32nd ACM International Conference on Multimedia , author =. 2024 , month = oct, pages =. doi:10.1145/3664647.3681648 , urldate =

work page doi:10.1145/3664647.3681648 2024
[16]

2024 , month = jul, journal =

Multi-Party Conversation Modeling for Emotion Recognition , author =. 2024 , month = jul, journal =. doi:10.1109/TAFFC.2023.3273589 , urldate =

work page doi:10.1109/taffc.2023.3273589 2024
[17]

Directed Acyclic Graph Network for Conversational Emotion Recognition

Shen, Weizhou and Wu, Siyue and Yang, Yunyi and Quan, Xiaojun , editor =. Directed Acyclic Graph Network for Conversational Emotion Recognition , booktitle =. 2021 , month = aug, pages =. doi:10.18653/v1/2021.acl-long.123 , urldate =

work page doi:10.18653/v1/2021.acl-long.123 2021
[18]

arXiv preprint arXiv:2407.00119 (2024)

Efficient Long-Distance Latent Relation-Aware Graph Neural Network for Multi-Modal Emotion Recognition in Conversations , author =. 2024 , month = aug, number =. 2407.00119 , publisher =

work page arXiv 2024
[19]

Narayanan

Busso, Carlos and Bulut, Murtaza and Lee, Chi-Chun and Kazemzadeh, Abe and Mower, Emily and Kim, Samuel and Chang, Jeannette N. and Lee, Sungbok and Narayanan, Shrikanth S. , year =. IEMOCAP: Interactive Emotional Dyadic Motion Capture Database , shorttitle =. Language Resources and Evaluation , volume =. doi:10.1007/s10579-008-9076-6 , urldate =

work page doi:10.1007/s10579-008-9076-6
[20]

Proceedings of the AAAI Conference on Artificial Intelligence , author =

Well, Now We Know! Unveiling Sarcasm: Initiating and Exploring Multimodal Conversations with Reasoning , shorttitle =. Proceedings of the AAAI Conference on Artificial Intelligence , author =. 2024 , month = mar, volume =. doi:10.1609/aaai.v38i17.29864 , urldate =

work page doi:10.1609/aaai.v38i17.29864 2024
[21]

2023 , journal =

SUNET: Speaker-Utterance Interaction Graph Neural Network for Emotion Recognition in Conversations , author =. 2023 , journal =

2023
[22]

Emotion Recognition in Conversation via Dynamic Personality , booktitle =

Wang, Yan and Wang, Bo and Zhao, Yachao and Zhao, Dongming and Jin, Xiaojia and Zhang, Jijun and He, Ruifang and Hou, Yuexian , editor =. Emotion Recognition in Conversation via Dynamic Personality , booktitle =. 2024 , month = may, pages =

2024
[23]

AIMDiT: Modality Augmentation and Interaction via Multimodal Dimension Transformation for Emotion Recognition in Conversations , shorttitle =

Wu, Sheng and Liu, Jiaxing and Wang, Longbiao and He, Dongxiao and Wang, Xiaobao and Dang, Jianwu , year =. AIMDiT: Modality Augmentation and Interaction via Multimodal Dimension Transformation for Emotion Recognition in Conversations , shorttitle =. 2407.00743 , publisher =

work page arXiv
[24]

Emotion-Anchored Contrastive Learning Framework for Emotion Recognition in Conversation , booktitle =

Yu, Fangxu and Guo, Junjie and Wu, Zhen and Dai, Xinyu , editor =. Emotion-Anchored Contrastive Learning Framework for Emotion Recognition in Conversation , booktitle =. 2024 , month = jun, pages =. doi:10.18653/v1/2024.findings-naacl.282 , urldate =

work page doi:10.18653/v1/2024.findings-naacl.282 2024
[25]

Proceedings of the Thirty-ThirdInternational Joint Conference on Artificial Intelligence , author =

A Survey on Network Alignment: Approaches, Applications and Future Directions , shorttitle =. Proceedings of the Thirty-ThirdInternational Joint Conference on Artificial Intelligence , author =. 2024 , month = aug, pages =. doi:10.24963/ijcai.2024/908 , urldate =

work page doi:10.24963/ijcai.2024/908 2024
[26]

2023 , publisher =

Comprehensive Evaluation of Unrestricted Graph Alignment Algorithms , author =. 2023 , publisher =. doi:10.48786/EDBT.2023.21 , urldate =

work page doi:10.48786/edbt.2023.21 2023
[27]

2020 , month = feb, journal =

A Comparative Study on Network Alignment Techniques , author =. 2020 , month = feb, journal =. doi:10.1016/j.eswa.2019.112883 , urldate =

work page doi:10.1016/j.eswa.2019.112883 2020
[28]

GOT: An Optimal Transport Framework for Graph Comparison , booktitle =

Maretic, Hermina Petric and Gheche, Mireille EL , year =. GOT: An Optimal Transport Framework for Graph Comparison , booktitle =
[29]

GALOPA: Graph Transport Learning with Optimal Plan Alignment , booktitle =

Wang, Yejiang and Zhao, Yuhai and Wang, Zhengkui and Li, Ling , year =. GALOPA: Graph Transport Learning with Optimal Plan Alignment , booktitle =
[30]

Scalable Gromov-Wasserstein Learning for Graph Partitioning and Matching , booktitle =

Xu, Hongteng and Luo, Dixin and Carin, Lawrence , year =. Scalable Gromov-Wasserstein Learning for Graph Partitioning and Matching , booktitle =
[31]

Proceedings of the ACM Web Conference 2023 , author =

PARROT: Position-Aware Regularized Optimal Transport for Network Alignment , shorttitle =. Proceedings of the ACM Web Conference 2023 , author =. 2023 , month = apr, pages =. doi:10.1145/3543507.3583357 , urldate =

work page doi:10.1145/3543507.3583357 2023
[32]

2024 , month = mar, journal =

Hierarchical Multi-Marginal Optimal Transport for Network Alignment , author =. 2024 , month = mar, journal =. doi:10.1609/aaai.v38i15.29605 , urldate =

work page doi:10.1609/aaai.v38i15.29605 2024
[33]

A Fused G romov- W asserstein Framework for Unsupervised Knowledge Graph Entity Alignment

Tang, Jianheng and Zhao, Kangfei and Li, Jia. A Fused G romov- W asserstein Framework for Unsupervised Knowledge Graph Entity Alignment. Findings of the Association for Computational Linguistics: ACL 2023. 2023. doi:10.18653/v1/2023.findings-acl.205

work page doi:10.18653/v1/2023.findings-acl.205 2023
[34]

Proceedings of the Seventh International Conference on Language Resources and Evaluation (LREC'10) , month = may, year =

Twitter as a Corpus for Sentiment Analysis and Opinion Mining , author =. Proceedings of the Seventh International Conference on Language Resources and Evaluation (LREC'10) , month = may, year =
[35]

2023 , journal =

Global-View and Speaker-Aware Emotion Cause Extraction in Conversations , author =. 2023 , journal =. doi:10.1109/TASLP.2023.3319990 , urldate =

work page doi:10.1109/taslp.2023.3319990 2023
[36]

SenticNet 5: Discovering Conceptual Primitives for Sentiment Analysis by Means of Context Embeddings , shorttitle =

Cambria, Erik and Poria, Soujanya and Hazarika, Devamanyu and Kwok, Kenneth , year =. SenticNet 5: Discovering Conceptual Primitives for Sentiment Analysis by Means of Context Embeddings , shorttitle =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. doi:10.1609/aaai.v32i1.11559 , urldate =

work page doi:10.1609/aaai.v32i1.11559
[37]

Emotion Cause Detection with Linguistic Constructions , booktitle =

Chen, Ying and Lee, Sophia Yat Mei and Li, Shoushan and Huang, Chu-Ren , editor =. Emotion Cause Detection with Linguistic Constructions , booktitle =. 2010 , month = aug, pages =

2010
[38]

Conversational Emotion-Cause Pair Extraction with Guided Mixture of Experts , booktitle =

Jeong, DongJin and Bak, JinYeong , editor =. Conversational Emotion-Cause Pair Extraction with Guided Mixture of Experts , booktitle =. 2023 , month = may, pages =. doi:10.18653/v1/2023.eacl-main.240 , urldate =

work page doi:10.18653/v1/2023.eacl-main.240 2023
[39]

2022 , month = oct, number =

Discourse-Aware Emotion Cause Extraction in Conversations , author =. 2022 , month = oct, number =. doi:10.48550/arXiv.2210.14419 , urldate =. 2210.14419 , primaryclass =

work page doi:10.48550/arxiv.2210.14419 2022
[40]

Enhancing Emotion--Cause Pair Extraction in Conversation With Contextual Information , author =
[41]

Towards Deep Conversational Recommendations , booktitle =

Li, Raymond and Ebrahimi Kahou, Samira and Schulz, Hannes and Michalski, Vincent and Charlin, Laurent and Pal, Chris , year =. Towards Deep Conversational Recommendations , booktitle =
[42]

ECPEC: Emotion-Cause Pair Extraction in Conversations , shorttitle =

Li, Wei and Li, Yang and Pandelea, Vlad and Ge, Mengshi and Zhu, Luyao and Cambria, Erik , year =. ECPEC: Emotion-Cause Pair Extraction in Conversations , shorttitle =. IEEE Transactions on Affective Computing , volume =. doi:10.1109/TAFFC.2022.3216551 , urldate =

work page doi:10.1109/taffc.2022.3216551 2022
[44]

Emotion-Cause Pair Extraction in Conversations Based on Multi-Turn MRC with Position-Aware GCN , booktitle =

Liu, Chen and Niu, Changyong and Xie, Jinge and Jia, Yuxiang and Zan, Hongying , year =. Emotion-Cause Pair Extraction in Conversations Based on Multi-Turn MRC with Position-Aware GCN , booktitle =. doi:10.1109/IALP61005.2023.10337182 , urldate =

work page doi:10.1109/ialp61005.2023.10337182 2023
[45]

2021 , month = sep, journal =

Recognizing Emotion Cause in Conversations , author =. 2021 , month = sep, journal =. doi:10.1007/s12559-021-09925-7 , urldate =

work page doi:10.1007/s12559-021-09925-7 2021
[47]

2023 , month = jul, journal =

Multimodal Emotion-Cause Pair Extraction in Conversations , author =. 2023 , month = jul, journal =. doi:10.1109/TAFFC.2022.3226559 , urldate =

work page doi:10.1109/taffc.2022.3226559 2023
[48]

Enhancing Emotion-Cause Pair Extraction in Conversations via Center Event Detection and Reasoning , booktitle =

Wang, Botao and Tang, Keke and Zhu, Peican , editor =. Enhancing Emotion-Cause Pair Extraction in Conversations via Center Event Detection and Reasoning , booktitle =. 2024 , month = nov, pages =. doi:10.18653/v1/2024.findings-emnlp.632 , urldate =

work page doi:10.18653/v1/2024.findings-emnlp.632 2024
[49]

Scientific Reports , volume =

MULTICAUSENET temporal attention for multimodal emotion cause pair extraction , author =. Scientific Reports , volume =. 2025 , doi =

2025
[50]

Enhanced Generative Framework With LLMs for Multimodal Emotion-Cause Pair Extraction in Conversations , year=

Ju, Xincheng and Zhang, Dong and Li, Junhui and Li, Shoushan and Zhou, Guodong , journal=. Enhanced Generative Framework With LLMs for Multimodal Emotion-Cause Pair Extraction in Conversations , year=
[51]

2010 , month = may, publisher =

Twitter as a Corpus for Sentiment Analysis and Opinion Mining , booktitle =. 2010 , month = may, publisher =

2010
[52]

End-to-End Emotion-Cause Pair Extraction with Graph Convolutional Network , booktitle =

Chen, Ying and Hou, Wenjun and Li, Shoushan and Wu, Caicong and Zhang, Xiaoqiang , editor =. End-to-End Emotion-Cause Pair Extraction with Graph Convolutional Network , booktitle =. 2020 , month = dec, pages =. doi:10.18653/v1/2020.coling-main.17 , url =

work page doi:10.18653/v1/2020.coling-main.17 2020
[53]

Conditional Causal Relationships between Emotions and Causes in Texts , booktitle =

Chen, Xinhong and Li, Qing and Wang, Jianping , editor =. Conditional Causal Relationships between Emotions and Causes in Texts , booktitle =. 2020 , month = nov, pages =. doi:10.18653/v1/2020.emnlp-main.252 , url =

work page doi:10.18653/v1/2020.emnlp-main.252 2020
[54]

Joint Alignment of Multi-Task Feature and Label Spaces for Emotion Cause Pair Extraction , booktitle =

Chen, Shunjie and Shi, Xiaochuan and Li, Jingye and Wu, Shengqiong and Fei, Hao and Li, Fei and Ji, Donghong , editor =. Joint Alignment of Multi-Task Feature and Label Spaces for Emotion Cause Pair Extraction , booktitle =. 2022 , month = oct, pages =

2022
[55]

End-to-End Emotion-Cause Pair Extraction Based on Sliding Window Multi-Label Learning , booktitle =

Ding, Zixiang and Xia, Rui and Yu, Jianfei , editor =. End-to-End Emotion-Cause Pair Extraction Based on Sliding Window Multi-Label Learning , booktitle =. 2020 , month = nov, pages =. doi:10.18653/v1/2020.emnlp-main.290 , url =

work page doi:10.18653/v1/2020.emnlp-main.290 2020
[56]

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , author =

ECPE-2D: Emotion-Cause Pair Extraction Based on Joint Two-Dimensional Representation, Interaction and Prediction , shorttitle =. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , author =. 2020 , month = jul, pages =. doi:10.18653/v1/2020.acl-main.288 , url =

work page doi:10.18653/v1/2020.acl-main.288 2020
[57]

2021 , journal =

Multi-Task Sequence Tagging for Emotion-Cause Pair Extraction Via Tag Distribution Refinement , author =. 2021 , journal =. doi:10.1109/TASLP.2021.3089837 , url =

work page doi:10.1109/taslp.2021.3089837 2021
[58]

Event-Driven Emotion Cause Extraction with Corpus Construction , booktitle =

Gui, Lin and Wu, Dongyin and Xu, Ruifeng and Lu, Qin and Zhou, Yu , editor =. Event-Driven Emotion Cause Extraction with Corpus Construction , booktitle =. 2016 , month = nov, pages =. doi:10.18653/v1/D16-1170 , url =

work page doi:10.18653/v1/d16-1170 2016
[59]

A Text-Driven Rule-Based System for Emotion Cause Detection , booktitle =

Lee, Sophia Yat Mei and Chen, Ying and Huang, Chu-Ren , editor =. A Text-Driven Rule-Based System for Emotion Cause Detection , booktitle =. 2010 , month = jun, pages =

2010
[60]

2024/7/8/23 , journal =

Multimodal Emotion-Cause Pair Extraction with Holistic Interaction and Label Constraint , author =. 2024/7/8/23 , journal =. doi:10.1145/3689646 , url =

work page doi:10.1145/3689646 2024
[61]

Findings of the Association for Computational Linguistics: ACL 2024 , author =

LLM-REDIAL: A Large-Scale Dataset for Conversational Recommender Systems Created from User Behaviors with LLMs , shorttitle =. Findings of the Association for Computational Linguistics: ACL 2024 , author =. 2024 , month = aug, pages =. doi:10.18653/v1/2024.findings-acl.529 , url =

work page doi:10.18653/v1/2024.findings-acl.529 2024
[62]

2022 , publisher =

Sentiment Analysis and Opinion Mining , author =. 2022 , publisher =

2022
[63]

S em E val-2016 Task 5: Aspect Based Sentiment Analysis

SemEval-2016 Task 5: Aspect Based Sentiment Analysis , shorttitle =. Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016) , author =. 2016 , month = jun, pages =. doi:10.18653/v1/S16-1002 , url =

work page doi:10.18653/v1/s16-1002 2016
[64]

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics , pages =

Towards Empathetic Open-Domain Conversation Models: A New Benchmark and Dataset , shorttitle =. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics , author =. 2019 , month = jul, pages =. doi:10.18653/v1/P19-1534 , url =

work page doi:10.18653/v1/p19-1534 2019
[65]

2023 , journal =

Recent Trends in Deep Learning Based Textual Emotion Cause Extraction , author =. 2023 , journal =. doi:10.1109/TASLP.2023.3254166 , url =

work page doi:10.1109/taslp.2023.3254166 2023
[66]

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics , author =

Emotion-Cause Pair Extraction: A New Task to Emotion Analysis in Texts , shorttitle =. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics , author =. 2019 , month = jul, pages =. doi:10.18653/v1/P19-1096 , url =

work page doi:10.18653/v1/p19-1096 2019
[67]

Towards Persona-Based Empathetic Conversational Models , booktitle =

Zhong, Peixiang and Zhang, Chen and Wang, Hao and Liu, Yong and Miao, Chunyan , editor =. Towards Persona-Based Empathetic Conversational Models , booktitle =. 2020 , month = nov, pages =. doi:10.18653/v1/2020.emnlp-main.531 , url =

work page doi:10.18653/v1/2020.emnlp-main.531 2020
[68]

2024 , month = feb, journal =

A Knowledge-Guided Graph Attention Network for Emotion-Cause Pair Extraction , author =. 2024 , month = feb, journal =. doi:10.1016/j.knosys.2023.111342 , url =

work page doi:10.1016/j.knosys.2023.111342 2024
[69]

Transactions of the association for computational linguistics , volume=

Spanbert: Improving pre-training by representing and predicting spans , author=. Transactions of the association for computational linguistics , volume=. 2020 , publisher=

2020
[70]

2019 , eprint=

RoBERTa: A Robustly Optimized BERT Pretraining Approach , author=. 2019 , eprint=

2019
[71]

Decoupled Weight Decay Regularization

Decoupled weight decay regularization , author=. arXiv preprint arXiv:1711.05101 , year=

work page internal anchor Pith review Pith/arXiv arXiv
[72]

PyTorch: an imperative style, high-performance deep learning library , year =

Paszke, Adam and Gross, Sam and Massa, Francisco and Lerer, Adam and Bradbury, James and Chanan, Gregory and Killeen, Trevor and Lin, Zeming and Gimelshein, Natalia and Antiga, Luca and Desmaison, Alban and K\". PyTorch: an imperative style, high-performance deep learning library , year =. Proceedings of the 33rd International Conference on Neural Informa...
[73]

2018 , eprint=

Graph Attention Networks , author=. 2018 , eprint=

2018
[74]

International Conference on Learning Representations , year=

How Attentive are Graph Attention Networks? , author=. International Conference on Learning Representations , year=
[75]

Proceedings of the 27th International Conference on Neural Information Processing Systems - Volume 2 , pages =

Cuturi, Marco , title =. Proceedings of the 27th International Conference on Neural Information Processing Systems - Volume 2 , pages =. 2013 , address =

2013
[76]

Proceedings of the 36th International Conference on Machine Learning , pages =

Optimal Transport for structured data with application on graphs , author =. Proceedings of the 36th International Conference on Machine Learning , pages =. 2019 , editor =

2019