arxiv: 2605.01717 · v2 · submitted 2026-05-03 · 💻 cs.CL · cs.AI

Recognition: no theorem link

TCDA: Thread-Constrained Discourse-Aware Modeling for Conversational Sentiment Quadruple Analysis

Xinran Li, Xinze Che, Xiujuan Xu, Yifan Lyu, Zhiqi Huang

Authors on Pith no claims yet

Pith reviewed 2026-05-11 00:43 UTC · model grok-4.3

classification 💻 cs.CL cs.AI

keywords conversational sentiment analysisaspect-based sentiment quadrupledirected acyclic graphrotary position embeddingdialogue modelingthread constraints

0 comments

The pith

Thread-constrained directed acyclic graphs and discourse-aware position embeddings capture dialogue structure for sentiment quadruple analysis.

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

The paper aims to improve conversational aspect-based sentiment quadruple analysis by addressing limitations in existing graph-based and position embedding methods. It introduces a framework that uses thread constraints to reduce noise and maintain temporal order in dialogues. The approach also employs specialized position embeddings to handle multi-scale semantics and dependencies between utterances. If effective, this would allow more accurate extraction of sentiment quadruples like aspect, opinion, sentiment, and holder from complex multi-turn conversations.

Core claim

The central discovery is that combining a Thread-Constrained Directed Acyclic Graph (TC-DAG) with Discourse-Aware Rotary Position Embedding (D-RoPE) enables better modeling of interrelationships in dialogues by filtering cross-thread noise, incorporating temporal sequences, aligning multi-layer semantics, and alleviating distance dilution problems.

What carries the argument

Thread-Constrained Directed Acyclic Graph (TC-DAG) for noise filtering and temporal ordering, paired with Discourse-Aware Rotary Position Embedding (D-RoPE) for semantic alignment and dependency capture using tree-like distances.

Load-bearing premise

That the observed improvements are due to the specific design of TC-DAG and D-RoPE rather than implementation details or dataset characteristics.

What would settle it

An ablation study removing the thread constraints from the graph or the discourse-aware modifications from the position embeddings that results in no significant performance loss would challenge the contribution of these components.

Figures

Figures reproduced from arXiv: 2605.01717 by Xinran Li, Xinze Che, Xiujuan Xu, Yifan Lyu, Zhiqi Huang.

**Figure 2.** Figure 2: The overall architecture of our proposed TCDA. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Figure 3: TC-DAG construction (ω = 1). Solid/dashed arrows denote inter- and same-speaker dependencies among chronological utterances. The structure incorporates Global Root Accessibility, allowing nodes in divergent threads (e.g., u6, u7) to connect to the global root u1 under the window constraint. Let h (l) i represent the hidden state of utterance ui in the lth layer, where the input state h (0) i correspond… view at source ↗

read the original abstract

Conversational Aspect-based Sentiment Quadruple Analysis (DiaASQ) needs to capture the complex interrelationships in multiple rounds of dialogues. Existing methods usually employ simple Graph Convolutional Networks (GCN), which introduce structural noise and fail to consider the temporal sequence of the dialogues, or use standard RoPE, which implicitly captures relative distances in a flat sequence but cannot clearly separate the token-level syntactic order from the utterance-level progression, and may suffer from the Distance Dilution problem. To address these issues, we propose a new framework that combines Thread-Constrained Directed Acyclic Graph (TC-DAG) and Discourse-Aware Rotary Position Embedding (D-RoPE). Specifically, TC-DAG filters out cross-thread noise based on thread constraints, maintains global connectivity through root anchoring, and incorporates the temporal sequence of the dialogues. D-RoPE aligns multi-layer semantics using dual-stream projection and multi-scale frequency signals, captures thread dependencies using tree-like distances, and alleviates the token-level Distance Dilution problem by incorporating utterance-level progressions. Experimental results on two benchmark datasets demonstrate that our framework achieves state-of-the-art performance.

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

TCDA pairs thread-constrained DAG filtering with a custom discourse-aware RoPE to cut cross-thread noise and distance dilution in DiaASQ, but the SOTA claim lacks ablations to pin gains on those pieces.

read the letter

The paper's core move is to replace plain GCNs with a TC-DAG that respects thread boundaries while keeping root-level connectivity, and to replace standard RoPE with D-RoPE that adds dual-stream projections, tree-like distances, and utterance-level signals. This directly targets two documented problems: structural noise from unrelated threads and the flattening of long dialogue sequences. The motivation is clear and the fixes are specific rather than generic graph or position tweaks. That combination is the actual novelty here. The design choices read as thoughtful engineering for the conversational setting. Root anchoring and multi-scale frequencies are reasonable ways to preserve both local syntax and global thread flow. The two DiaASQ benchmarks are the standard ones, so the evaluation setup is at least on the right datasets. The main soft spot is attribution. The abstract states SOTA results, yet the stress-test note and available description give no sign of ablations that isolate TC-DAG or D-RoPE from baseline GCN+RoPE runs, nor any multi-seed stats or error bars. Without those, the gains could come from tuning, data quirks, or unstated implementation details rather than the proposed mechanisms. Two benchmarks also leave open how well it holds on noisier real dialogues. This is for researchers working on aspect-based sentiment in multi-turn conversations or dialogue systems. A reader already familiar with GCN and RoPE variants will see the targeted extensions and can judge whether the modeling ideas are worth trying. It deserves peer review. The problem is well scoped, the architecture is concrete, and referees can demand the missing controls and statistical checks. Send it out rather than desk reject.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces the TCDA framework for Conversational Aspect-based Sentiment Quadruple Analysis (DiaASQ). It proposes Thread-Constrained Directed Acyclic Graph (TC-DAG) to filter cross-thread noise while maintaining global connectivity and incorporating temporal dialogue sequences, and Discourse-Aware Rotary Position Embedding (D-RoPE) to align semantics with dual-stream projection, capture thread dependencies via tree-like distances, and mitigate distance dilution using utterance-level signals. The authors report that this framework achieves state-of-the-art performance on two benchmark datasets.

Significance. Should the empirical superiority be confirmed through rigorous controls, this work would contribute to the field by offering a structured approach to modeling discourse and thread constraints in conversational sentiment analysis, potentially improving accuracy in tasks involving multi-turn dialogues.

major comments (2)

[Experimental Results] Experimental Results: The central SOTA claim lacks supporting ablation experiments that isolate the effects of TC-DAG (e.g., without thread constraints or root anchoring) and D-RoPE (e.g., replaced by standard RoPE). Without these comparisons, it is not possible to attribute performance gains specifically to the proposed components rather than baseline GCN+RoPE variants, hyperparameter choices, or other implementation details.
[Methods] Methods: The descriptions of TC-DAG and D-RoPE remain at a high level without formal mathematical definitions, pseudocode, or explicit equations for key mechanisms such as root anchoring, dual-stream projection, or multi-scale frequency signals, hindering verification of the claimed properties.

minor comments (1)

[Abstract] The abstract would be strengthened by including specific performance metrics (e.g., F1 scores) and the names of the two benchmark datasets to provide immediate evidence for the SOTA claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and outline the revisions we will make to strengthen the empirical validation and methodological clarity.

read point-by-point responses

Referee: [Experimental Results] Experimental Results: The central SOTA claim lacks supporting ablation experiments that isolate the effects of TC-DAG (e.g., without thread constraints or root anchoring) and D-RoPE (e.g., replaced by standard RoPE). Without these comparisons, it is not possible to attribute performance gains specifically to the proposed components rather than baseline GCN+RoPE variants, hyperparameter choices, or other implementation details.

Authors: We agree that additional ablation studies are required to rigorously isolate the contributions of TC-DAG and D-RoPE. In the revised manuscript, we will include targeted ablations such as TC-DAG without thread constraints or root anchoring, and D-RoPE replaced by standard RoPE, along with comparisons to baseline GCN+RoPE variants. These experiments will be reported with performance metrics on both benchmarks to attribute gains specifically to the proposed components. revision: yes
Referee: [Methods] Methods: The descriptions of TC-DAG and D-RoPE remain at a high level without formal mathematical definitions, pseudocode, or explicit equations for key mechanisms such as root anchoring, dual-stream projection, or multi-scale frequency signals, hindering verification of the claimed properties.

Authors: We acknowledge that the current high-level descriptions limit verifiability. We will expand the Methods section with formal mathematical definitions, including explicit equations for root anchoring in TC-DAG, dual-stream projection, tree-like distances, and multi-scale frequency signals in D-RoPE, as well as pseudocode for the core mechanisms to support the claimed properties and improve reproducibility. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical SOTA claims rest on benchmark experiments, not derivations

full rationale

The paper proposes TC-DAG and D-RoPE components to address limitations in GCN and standard RoPE for conversational sentiment quadruple analysis. Its strongest claims are that the combined framework achieves state-of-the-art performance on two DiaASQ benchmark datasets. No equations, first-principles derivations, or mathematical predictions are presented anywhere in the provided text. There are no self-definitional steps, fitted parameters renamed as predictions, or load-bearing self-citations that reduce the results to inputs by construction. The performance improvements are framed purely as empirical outcomes from experiments, making the derivation chain self-contained against external benchmarks with no reduction to tautology.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 2 invented entities

Based on abstract only. The central claim rests on the unproven premise that the new structures solve the stated problems without side effects.

axioms (2)

domain assumption Simple GCNs introduce structural noise and ignore temporal sequence in dialogues
Invoked as motivation without supporting analysis in the abstract
domain assumption Standard RoPE suffers from Distance Dilution and cannot separate token-level and utterance-level order
Stated as a limitation without quantitative demonstration

invented entities (2)

TC-DAG no independent evidence
purpose: Filter cross-thread noise, maintain global connectivity via root anchoring, and incorporate temporal sequence
Newly introduced graph structure
D-RoPE no independent evidence
purpose: Align multi-layer semantics with dual-stream projection and multi-scale frequencies while capturing thread dependencies
Newly introduced position embedding variant

pith-pipeline@v0.9.0 · 5510 in / 1408 out tokens · 55366 ms · 2026-05-11T00:43:13.829241+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Reference graph

Works this paper leans on

42 extracted references · 28 canonical work pages

[1]

D ia ASQ : A Benchmark of Conversational Aspect-based Sentiment Quadruple Analysis

Li, Bobo and Fei, Hao and Li, Fei and Wu, Yuhan and Zhang, Jinsong and Wu, Shengqiong and Li, Jingye and Liu, Yijiang and Liao, Lizi and Chua, Tat-Seng and Ji, Donghong. D ia ASQ : A Benchmark of Conversational Aspect-based Sentiment Quadruple Analysis. Findings of the Association for Computational Linguistics: ACL 2023. 2023. doi:10.18653/v1/2023.finding...

work page doi:10.18653/v1/2023.findings-acl.849 2023
[2]

Context-Aware Directed Acyclic Graph Network for Conversational Aspect-Based Sentiment Quadruple Analysis , year=

Zhang, Qiang and Zeng, Jie and Zhang, Runze and Cui, Dong , journal=. Context-Aware Directed Acyclic Graph Network for Conversational Aspect-Based Sentiment Quadruple Analysis , year=
[3]

DMIN : A Discourse-specific Multi-granularity Integration Network for Conversational Aspect-based Sentiment Quadruple Analysis

Huang, Peijie and Xiao, Xisheng and Xu, Yuhong and Chen, Jiawei. DMIN : A Discourse-specific Multi-granularity Integration Network for Conversational Aspect-based Sentiment Quadruple Analysis. Findings of the Association for Computational Linguistics: ACL 2024. 2024. doi:10.18653/v1/2024.findings-acl.966

work page doi:10.18653/v1/2024.findings-acl.966 2024
[4]

Multi-level Association Refinement Network for Dialogue Aspect-based Sentiment Quadruple Analysis

Tong, Zeliang and Wei, Wei and Qu, Xiaoye and Huang, Rikui and Chen, Zhixin and Yan, Xingyu. Multi-level Association Refinement Network for Dialogue Aspect-based Sentiment Quadruple Analysis. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025. doi:10.18653/v1/2025.acl-long.686

work page doi:10.18653/v1/2025.acl-long.686 2025
[5]

Li, Bobo and Fei, Hao and Liao, Lizi and Zhao, Yu and Su, Fangfang and Li, Fei and Ji, Donghong , title =. Proceedings of the Thirty-Eighth AAAI Conference on Artificial Intelligence and Thirty-Sixth Conference on Innovative Applications of Artificial Intelligence and Fourteenth Symposium on Educational Advances in Artificial Intelligence , articleno =. 2...

work page doi:10.1609/aaai.v38i16.29807 2024
[6]

and Bloem, Peter and van den Berg, Rianne and Titov, Ivan and Welling, Max

Schlichtkrull, Michael and Kipf, Thomas N. and Bloem, Peter and van den Berg, Rianne and Titov, Ivan and Welling, Max. Modeling Relational Data with Graph Convolutional Networks. The Semantic Web. 2018

2018
[7]

Dynamic Multi-Scale Context Aggregation for Conversational Aspect-Based Sentiment Quadruple Analysis , year=

Li, Yuqing and Zhang, Wenyuan and Li, Binbin and Jia, Siyu and Qi, Zisen and Tan, Xingbang , booktitle=. Dynamic Multi-Scale Context Aggregation for Conversational Aspect-Based Sentiment Quadruple Analysis , year=
[8]

Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =

Vedula, Nikhita and Collins, Marcus and Rokhlenko, Oleg , title =. Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =. 2023 , isbn =. doi:10.1145/3539618.3591974 , abstract =

work page doi:10.1145/3539618.3591974 2023
[9]

RoFormer: Enhanced Transformer with Rotary Position Embedding,

Jianlin Su and Murtadha Ahmed and Yu Lu and Shengfeng Pan and Wen Bo and Yunfeng Liu , keywords =. RoFormer: Enhanced transformer with Rotary Position Embedding , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.neucom.2023.127063 , url =

work page doi:10.1016/j.neucom.2023.127063 2024
[10]

H i R o PE : Length Extrapolation for Code Models Using Hierarchical Position

Zhang, Kechi and Li, Ge and Zhang, Huangzhao and Jin, Zhi. H i R o PE : Length Extrapolation for Code Models Using Hierarchical Position. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024. doi:10.18653/v1/2024.acl-long.735

work page doi:10.18653/v1/2024.acl-long.735 2024
[11]

In Proceedings of the 26th Annual International Conference on Machine Learning (Montreal, Quebec, Canada) (ICML ’09)

Bengio, Yoshua and Louradour, J\'. Curriculum learning , year =. Proceedings of the 26th Annual International Conference on Machine Learning , pages =. doi:10.1145/1553374.1553380 , abstract =

work page doi:10.1145/1553374.1553380
[12]

Proceedings of the 28th European Conference on Artificial Intelligence (ECAI 2025) , year =

Xinran Li and Xiujuan Xu and Jiaqi Qiao , title =. Proceedings of the 28th European Conference on Artificial Intelligence (ECAI 2025) , year =. doi:10.3233/FAIA251292 , series =

work page doi:10.3233/faia251292 2025
[13]

Proceedings of the AAAI Conference on Artificial Intelligence , author=

Do LLMs Feel? Teaching Emotion Recognition with Prompts, Retrieval, and Curriculum Learning , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2026 , month=. doi:10.1609/aaai.v40i38.40446 , number=

work page doi:10.1609/aaai.v40i38.40446 2026
[14]

Aspect Sentiment Quad Prediction as Paraphrase Generation

Zhang, Wenxuan and Deng, Yang and Li, Xin and Yuan, Yifei and Bing, Lidong and Lam, Wai. Aspect Sentiment Quad Prediction as Paraphrase Generation. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. 2021. doi:10.18653/v1/2021.emnlp-main.726

work page doi:10.18653/v1/2021.emnlp-main.726 2021
[15]

S eq2 P ath: Generating Sentiment Tuples as Paths of a Tree

Mao, Yue and Shen, Yi and Yang, Jingchao and Zhu, Xiaoying and Cai, Longjun. S eq2 P ath: Generating Sentiment Tuples as Paths of a Tree. Findings of the Association for Computational Linguistics: ACL 2022. 2022. doi:10.18653/v1/2022.findings-acl.174

work page doi:10.18653/v1/2022.findings-acl.174 2022
[16]

Dual Encoder: Exploiting the Potential of Syntactic and Semantic for Aspect Sentiment Triplet Extraction

Zhao, Xiaowei and Zhou, Yong and Xu, Xiujuan. Dual Encoder: Exploiting the Potential of Syntactic and Semantic for Aspect Sentiment Triplet Extraction. Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024). 2024

2024
[17]

Graph attention networks , url=

Veli. Graph attention networks , url=. doi:10.17863/CAM.48429 , publisher=

work page doi:10.17863/cam.48429
[18]

S em E val-2014 Task 4: Aspect Based Sentiment Analysis

Pontiki, Maria and Galanis, Dimitris and Pavlopoulos, John and Papageorgiou, Harris and Androutsopoulos, Ion and Manandhar, Suresh. S em E val-2014 Task 4: Aspect Based Sentiment Analysis. Proceedings of the 8th International Workshop on Semantic Evaluation ( S em E val 2014). 2014. doi:10.3115/v1/S14-2004

work page doi:10.3115/v1/s14-2004 2014
[19]

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

Pontiki, Maria and Galanis, Dimitris and Papageorgiou, Haris and Androutsopoulos, Ion and Manandhar, Suresh and AL-Smadi, Mohammad and Al-Ayyoub, Mahmoud and Zhao, Yanyan and Qin, Bing and De Clercq, Orph. S em E val-2016 Task 5: Aspect Based Sentiment Analysis. Proceedings of the 10th International Workshop on Semantic Evaluation ( S em E val-2016). 2016...

work page doi:10.18653/v1/s16-1002 2016
[20]

A Survey on Aspect-Based Sentiment Analysis: Tasks, Methods, and Challenges , year=

Zhang, Wenxuan and Li, Xin and Deng, Yang and Bing, Lidong and Lam, Wai , journal=. A Survey on Aspect-Based Sentiment Analysis: Tasks, Methods, and Challenges , year=
[21]

Proceedings of the ACM Web Conference 2024 , pages =

Zhou, Changzhi and Wu, Zhijing and Song, Dandan and Hu, Linmei and Tian, Yuhang and Xu, Jing , title =. Proceedings of the ACM Web Conference 2024 , pages =. 2024 , isbn =. doi:10.1145/3589334.3645355 , abstract =

work page doi:10.1145/3589334.3645355 2024
[22]

International Conference on Learning Representations , year=

Directed Acyclic Graph Neural Networks , author=. International Conference on Learning Representations , year=
[23]

Directed Acyclic Graph Network for Conversational Emotion Recognition

Shen, Weizhou and Wu, Siyue and Yang, Yunyi and Quan, Xiaojun. Directed Acyclic Graph Network for Conversational Emotion Recognition. 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). 2021. doi:10.18653/v1/2021.acl-long.123

work page doi:10.18653/v1/2021.acl-long.123 2021
[24]

Proceedings of the AAAI Conference on Artificial Intelligence , author=

Hybrid Curriculum Learning for Emotion Recognition in Conversation , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2022 , month=. doi:10.1609/aaai.v36i10.21413 , abstractNote=

work page doi:10.1609/aaai.v36i10.21413 2022
[25]

Curriculum Learning Meets Directed Acyclic Graph for Multimodal Emotion Recognition

Nguyen, Cam-Van Thi and Nguyen, Cao-Bach and Le, Duc-Trong and Ha, Quang-Thuy. Curriculum Learning Meets Directed Acyclic Graph for Multimodal Emotion Recognition. Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024). 2024

2024
[26]

BERT: Pre-training of deep bidi- rectional transformers for language understanding

Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina. BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding. Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). 2019. doi:10.18653/v...

work page doi:10.18653/v1/n19-1423 2019
[27]

ArXiv , year=

Semi-Supervised Classification with Graph Convolutional Networks , author=. ArXiv , year=
[28]

Enhanced Multi-Channel Graph Convolutional Network for Aspect Sentiment Triplet Extraction

Chen, Hao and Zhai, Zepeng and Feng, Fangxiang and Li, Ruifan and Wang, Xiaojie. Enhanced Multi-Channel Graph Convolutional Network for Aspect Sentiment Triplet Extraction. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022. doi:10.18653/v1/2022.acl-long.212

work page doi:10.18653/v1/2022.acl-long.212 2022
[29]

SSEGCN : Syntactic and Semantic Enhanced Graph Convolutional Network for Aspect-based Sentiment Analysis

Zhang, Zheng and Zhou, Zili and Wang, Yanna. SSEGCN : Syntactic and Semantic Enhanced Graph Convolutional Network for Aspect-based Sentiment Analysis. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022. doi:10.18653/v1/2022.naacl-main.362

work page doi:10.18653/v1/2022.naacl-main.362 2022
[30]

Neural Information Processing Systems , year=

Attention is All you Need , author=. Neural Information Processing Systems , year=
[31]

International Joint Conference on Artificial Intelligence , year=

Aspect Term Extraction with History Attention and Selective Transformation , author=. International Joint Conference on Artificial Intelligence , year=
[32]

Dual Graph Convolutional Networks for Aspect-based Sentiment Analysis

Li, Ruifan and Chen, Hao and Feng, Fangxiang and Ma, Zhanyu and Wang, Xiaojie and Hovy, Eduard. Dual Graph Convolutional Networks for Aspect-based Sentiment Analysis. 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). 20...

work page doi:10.18653/v1/2021.acl-long.494 2021
[33]

2024 , url=

Extensible Multi-Granularity Fusion Network and Transferable Curriculum Learning for Aspect-based Sentiment Analysis , author=. 2024 , url=

2024
[34]

International Joint Conference on Artificial Intelligence , year=

Learn from Syntax: Improving Pair-wise Aspect and Opinion Terms Extractionwith Rich Syntactic Knowledge , author=. International Joint Conference on Artificial Intelligence , year=
[35]

Aspect-Category-Opinion-Sentiment Quadruple Extraction with Implicit Aspects and Opinions

Cai, Hongjie and Xia, Rui and Yu, Jianfei. Aspect-Category-Opinion-Sentiment Quadruple Extraction with Implicit Aspects and Opinions. 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). 2021. doi:10.18653/v1/2021.acl-long.29

work page doi:10.18653/v1/2021.acl-long.29 2021
[36]

ArXiv , year=

RoBERTa: A Robustly Optimized BERT Pretraining Approach , author=. ArXiv , year=
[37]

IEEE/ACM Transactions on Audio, Speech, and Language Processing , year=

Pre-Training With Whole Word Masking for Chinese BERT , author=. IEEE/ACM Transactions on Audio, Speech, and Language Processing , year=
[38]

Learning Span-Level Interactions for Aspect Sentiment Triplet Extraction

Xu, Lu and Chia, Yew Ken and Bing, Lidong. Learning Span-Level Interactions for Aspect Sentiment Triplet Extraction. 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). 2021. doi:10.18653/v1/2021.acl-long.367

work page doi:10.18653/v1/2021.acl-long.367 2021
[39]

Natural Language Processing and Chinese Computing: 12th National CCF Conference, NLPCC 2023, Foshan, China, October 12–15, 2023, Proceedings, Part III , pages =

Cai, Chenran and Zhao, Qin and Xu, Ruifeng and Qin, Bing , title =. Natural Language Processing and Chinese Computing: 12th National CCF Conference, NLPCC 2023, Foshan, China, October 12–15, 2023, Proceedings, Part III , pages =. 2023 , isbn =. doi:10.1007/978-3-031-44699-3_14 , abstract =

work page doi:10.1007/978-3-031-44699-3_14 2023
[40]

Inter-sentence Context Modeling and Structure-aware Representation Enhancement for Conversational Sentiment Quadruple Extraction

Zhang, Yu and Zhong, Zhaoman and Lv, Huihui. Inter-sentence Context Modeling and Structure-aware Representation Enhancement for Conversational Sentiment Quadruple Extraction. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025. doi:10.18653/v1/2025.emnlp-main.867

work page doi:10.18653/v1/2025.emnlp-main.867 2025
[41]

IFusionQuad: A novel framework for improved aspect-based sentiment quadruple analysis in dialogue contexts with advanced feature integration and contextual CloBlock , journal =

Haoyu Jiang and Xiaoliang Chen and Duoqian Miao and Hongyun Zhang and Xiaolin Qin and Xu Gu and Peng Lu , keywords =. IFusionQuad: A novel framework for improved aspect-based sentiment quadruple analysis in dialogue contexts with advanced feature integration and contextual CloBlock , journal =. 2025 , issn =. doi:https://doi.org/10.1016/j.eswa.2024.125556 , url =

work page doi:10.1016/j.eswa.2024.125556 2025
[42]

SARA: Span-aware framework with relation-augmented grid tagging for conversational aspect-based sentiment quadruple analysis , journal =

Xiaoyong Liu and Miao Hu and Chunlin Xu and Zhiguo Du , keywords =. SARA: Span-aware framework with relation-augmented grid tagging for conversational aspect-based sentiment quadruple analysis , journal =. 2025 , issn =. doi:https://doi.org/10.1016/j.engappai.2025.110915 , url =

work page doi:10.1016/j.engappai.2025.110915 2025