pith. sign in

arxiv: 2605.22228 · v1 · pith:M66YFTEZnew · submitted 2026-05-21 · 💻 cs.CL

GHI: Graphormer over Conditioned Hypergraph Incidence for Aspect-Based Sentiment Analysis

Yu Du , Wenlong Zhu , Xingze Li , Chenglong Cao , Jing Wang , Yukun Ma This is my paper

Pith reviewed 2026-05-22 06:03 UTC · model grok-4.3

classification 💻 cs.CL
keywords Aspect-Based Sentiment AnalysisHypergraph IncidenceGraphormerStructural ReasoningBipartite TopologySentiment AnalysisNLP Models
0
0 comments X

The pith

GHI shows that representing linguistic evidence as token-hyperedge incidences on a bipartite topology lets a compact model rival much larger systems on aspect-based sentiment analysis.

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

The paper introduces GHI as a Graphormer-based layer that treats ABSA as a problem of binding sentiment evidence to aspects through explicit structural relations. It claims this incidence representation unifies different linguistic signals and yields gains without depending on extreme scale. A sympathetic reader would care because the results indicate that such compact structural methods can match or approach 11B-parameter approaches on standard benchmarks while holding up better under robustness tests.

Core claim

GHI is an incidence-based structural reasoning layer built on a bipartite topology. It represents diverse linguistic and semantic evidence as token-hyperedge incidence relations, allowing different structural signals to be incorporated through a unified interface. Extensive experiments on six standard ABSA benchmarks show that GHI outperforms all baselines on the SemEval domains, delivers stable improvements over strong DeBERTa, approaches the performance of 11B Flan-T5 based methods on the ISE benchmark with only 247M parameters, and demonstrates strong robustness on the challenging ARTS datasets where traditional models degrade.

What carries the argument

The conditioned hypergraph incidence structure, which serves as the bipartite topology enabling Graphormer to process token-hyperedge relations as a unified interface for structural signals.

If this is right

  • Outperforms all baselines on the SemEval domains.
  • Shows stable improvements over strong DeBERTa in multi-seed evaluations.
  • Approaches 11B Flan-T5 performance on the ISE benchmark using only 247M parameters.
  • Maintains highly competitive performance on ARTS datasets where traditional models degrade.

Where Pith is reading between the lines

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

  • The same incidence mechanism might transfer to other fine-grained tasks that require binding evidence across relations.
  • Treating structure as a lightweight alternative could lower the compute needed for specialized NLP applications.
  • Varying the conditioning on hyperedges might reveal which structural signals contribute most to the observed robustness.

Load-bearing premise

That diverse linguistic and semantic evidence can be represented as token-hyperedge incidence relations in a way that allows different structural signals to be incorporated through a unified interface.

What would settle it

A direct comparison showing that GHI loses its robustness advantage on the ARTS datasets or that its performance gains vanish when the hypergraph incidence component is ablated would falsify the central claim.

Figures

Figures reproduced from arXiv: 2605.22228 by Chenglong Cao, Jing Wang, Wenlong Zhu, Xingze Li, Yu Du, Yukun Ma.

Figure 1
Figure 1. Figure 1: An example sentence with two different as [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: A hypergraph view for the aspect "price". [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The overall architecture of the proposed [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Distance distribution of adaptive Top-K to￾kens relative to aspect spans on SemEval-14 domains. Error bars denote 95% bootstrap confidence intervals. The food - - though mostly deep - fried - - is simple and satisfying . food Startup times are incredibly long : over two minutes . Startup times [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visualization examples in two cases that the learned incidence can capture multi-token and even implicit evidence expressions toward the target aspect. 4 Related Work Structural modeling remains important for ABSA, where models must bind sentiment evidence to the correct aspect. Recent methods refine syntactic, semantic, and aspect-specific structures from dif￾ferent perspectives. Early efforts in this dir… view at source ↗
Figure 6
Figure 6. Figure 6: Adaptive Incidence Generation Token States Token States Hyperedge States Token Update Attention over Incidences (conditioned by ) Token Nodes Hyperedge Nodes Inputs: Output: Local Incidence Reasoning Output: Local Pooling Star-Expanded Graphormer Inputs: Output: Fusion and State Update Anchor Memory Update Repeated for [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Layer-wise propagation in one GHI reasoning [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Sensitivity analyses for number of Adaptive Top- [PITH_FULL_IMAGE:figures/full_fig_p015_8.png] view at source ↗
read the original abstract

Aspect-based sentiment analysis (ABSA) requires models to bind sentiment evidence to the correct aspect, making it a natural testbed for fine-grained structural reasoning. We introduce GHI, a Graphormer-over-Conditioned-Hypergraph-Incidence framework that is designed as an incidence-based structural reasoning layer built on a bipartite topology. GHI represents diverse linguistic and semantic evidence as token--hyperedge incidence relations, allowing different structural signals to be incorporated through a unified interface. Extensive experiments on six standard ABSA benchmarks show that GHI outperforms all baselines on the SemEval domains, and multi-seed evaluations show stable improvements over strong DeBERTa. Further experiments show that with only 247M parameters, GHI approaches the performance of 11B Flan-T5 based methods on the ISE benchmark. Moreover, it demonstrates strong robustness on the challenging ARTS datasets, maintaining highly competitive performance where traditional models degrade. These results demonstrate that compact structural reasoning remains a valuable alternative to scale-driven approaches for fine-grained tasks.

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.

Referee Report

2 major / 2 minor

Summary. The paper introduces GHI, a Graphormer-over-Conditioned-Hypergraph-Incidence framework for aspect-based sentiment analysis. It models diverse linguistic and semantic evidence as token-hyperedge incidence relations on a bipartite topology to enable unified structural reasoning. Experiments on six ABSA benchmarks report outperformance over baselines on SemEval domains, stable gains over DeBERTa with multi-seed runs, competitive results to 11B Flan-T5 on ISE using only 247M parameters, and strong robustness on ARTS where other models degrade.

Significance. If the central claims hold after proper isolation of the incidence component, the work would demonstrate that compact, incidence-based structural reasoning can serve as an efficient alternative to scale-driven approaches for fine-grained sentiment tasks, with potential implications for parameter-efficient modeling in NLP.

major comments (2)
  1. [Experiments] Experiments section (around the ablation and comparison tables): The central claim that the conditioned hypergraph incidence representation drives the reported SemEval improvements and ARTS robustness requires isolation from the base encoder and Graphormer capacity. No ablation is described that removes or randomizes the incidence matrix while holding Graphormer layers, base parameters, and training schedule fixed; without this, performance differences could stem from capacity or schedule rather than the claimed structural interface.
  2. [Method] Method section (hyperedge construction and conditioning): The assumption that aspects, opinions, and syntactic signals can be represented as hyperedges with conditioning applied in a unified bipartite topology is load-bearing for the robustness claims. The manuscript does not provide a concrete, reproducible procedure for hyperedge definition or conditioning that would allow verification independent of parser heuristics.
minor comments (2)
  1. [Abstract] Abstract and §1: The phrase 'approaches the performance of 11B Flan-T5' should be accompanied by the exact metric values and the specific ISE benchmark split for direct comparison.
  2. [Figures/Tables] Figure captions and tables: Ensure all reported numbers include standard deviations from the multi-seed runs and clarify whether the 247M parameter count includes the base encoder or only the GHI layer.

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 improve clarity and strengthen the empirical support for our claims.

read point-by-point responses

  1. Referee: [Experiments] The central claim that the conditioned hypergraph incidence representation drives the reported SemEval improvements and ARTS robustness requires isolation from the base encoder and Graphormer capacity. No ablation is described that removes or randomizes the incidence matrix while holding Graphormer layers, base parameters, and training schedule fixed; without this, performance differences could stem from capacity or schedule rather than the claimed structural interface.

    Authors: We agree that isolating the incidence matrix contribution is important for validating the central claim. In the revised version, we will add a controlled ablation that replaces the learned incidence matrix with a randomized matrix of identical shape and sparsity while freezing the Graphormer architecture, base encoder weights, and training hyperparameters. Results from this ablation will be reported alongside the existing tables to demonstrate that performance gains are attributable to the structured incidence relations rather than incidental capacity differences. revision: yes

  2. Referee: [Method] The assumption that aspects, opinions, and syntactic signals can be represented as hyperedges with conditioning applied in a unified bipartite topology is load-bearing for the robustness claims. The manuscript does not provide a concrete, reproducible procedure for hyperedge definition or conditioning that would allow verification independent of parser heuristics.

    Authors: We acknowledge that the current description of hyperedge construction could be made more explicit for independent reproduction. Section 3.2 outlines the use of dependency parses to form hyperedges for aspect-opinion pairs and syntactic constituents, with conditioning implemented via masked attention in the bipartite incidence graph. In the revision, we will include a detailed algorithmic procedure with pseudocode, specify the exact input format expected from any dependency parser, and clarify how conditioning vectors are derived and applied, thereby reducing reliance on specific parser heuristics. revision: yes

Circularity Check

0 steps flagged

No circularity: architectural proposal with external empirical validation

full rationale

The paper introduces GHI as a new model architecture consisting of a Graphormer layer over conditioned hypergraph incidence on a bipartite token-hyperedge topology. No equations, parameter-fitting steps, or derivations are presented that reduce the claimed structural reasoning or performance gains to self-defined quantities or prior self-citations. The central claims rest on experimental results across SemEval benchmarks, ARTS robustness tests, and comparisons to DeBERTa and Flan-T5 baselines, which constitute independent external evidence rather than tautological reuse of the model's own inputs. The design choices (incidence relations for linguistic signals) are presented as an engineering construction, not as a result derived from the target metrics by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

Abstract-only review provides no explicit free parameters, background axioms, or additional invented entities beyond the model architecture itself; the ledger is therefore minimal.

invented entities (1)
  • Conditioned Hypergraph Incidence structure no independent evidence
    purpose: Represent linguistic and semantic evidence as token-hyperedge incidence relations on a bipartite topology to enable unified structural signal incorporation
    Core representational choice introduced by the GHI framework in the abstract

pith-pipeline@v0.9.0 · 5720 in / 1379 out tokens · 87563 ms · 2026-05-22T06:03:16.611932+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

50 extracted references · 50 canonical work pages · 1 internal anchor

  1. [1]

    Aho and Jeffrey D

    Alfred V. Aho and Jeffrey D. Ullman , title =. 1972

    work page 1972

  2. [2]

    Publications Manual , year = "1983", publisher =

    work page 1983

  3. [3]

    Chandra and Dexter C

    Ashok K. Chandra and Dexter C. Kozen and Larry J. Stockmeyer , year = "1981", title =. doi:10.1145/322234.322243

    work page doi:10.1145/322234.322243 1981

  4. [4]

    Scalable training of

    Andrew, Galen and Gao, Jianfeng , booktitle=. Scalable training of

    work page

  5. [5]

    Dan Gusfield , title =. 1997

    work page 1997

  6. [6]

    Tetreault , title =

    Mohammad Sadegh Rasooli and Joel R. Tetreault , title =. Computing Research Repository , volume =. 2015 , url =

    work page 2015

  7. [7]

    A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =

    Ando, Rie Kubota and Zhang, Tong , Issn =. A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =. Journal of Machine Learning Research , Month = dec, Numpages =

    work page

  8. [8]

    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=

    work page

  9. [9]

    DAGCN : Distance-based and Aspect-oriented Graph Convolutional Network for Aspect-based Sentiment Analysis

    Wang, Zhihao and Zhang, Bo and Yang, Ru and Guo, Chang and Li, Maozhen. DAGCN : Distance-based and Aspect-oriented Graph Convolutional Network for Aspect-based Sentiment Analysis. Findings of the Association for Computational Linguistics: NAACL 2024. 2024. doi:10.18653/v1/2024.findings-naacl.120

    work page doi:10.18653/v1/2024.findings-naacl.120 2024

  10. [10]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    TextGT: A Double-View Graph Transformer on Text for Aspect-Based Sentiment Analysis , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2024 , month=. doi:10.1609/aaai.v38i17.29911 , abstractNote=

    work page doi:10.1609/aaai.v38i17.29911 2024

  11. [11]

    Aspect-based sentiment classification with aspect-specific hypergraph attention networks , journal =

    Jihong Ouyang and Chang Xuan and Bing Wang and Zhiyao Yang , keywords =. Aspect-based sentiment classification with aspect-specific hypergraph attention networks , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.eswa.2024.123412 , url =

    work page doi:10.1016/j.eswa.2024.123412 2024

  12. [12]

    Dual contrastive learning-based hypergraph convolutional network for aspect-based sentiment classification , journal =

    Xinyi Ju and Ling Ding and Ru Yang and Chang Guo and Guojian Zou and Bo Zhang and Meizi Li , keywords =. Dual contrastive learning-based hypergraph convolutional network for aspect-based sentiment classification , journal =. 2025 , issn =. doi:https://doi.org/10.1016/j.knosys.2025.114701 , url =

    work page doi:10.1016/j.knosys.2025.114701 2025

  13. [13]

    2025 , eprint=

    From Graphs to Hypergraphs: Enhancing Aspect-Based Sentiment Analysis via Multi-Level Relational Modeling , author=. 2025 , eprint=

    work page 2025

  14. [14]

    Do Transformers Really Perform Badly for Graph Representation? , url =

    Ying, Chengxuan and Cai, Tianle and Luo, Shengjie and Zheng, Shuxin and Ke, Guolin and He, Di and Shen, Yanming and Liu, Tie-Yan , booktitle =. Do Transformers Really Perform Badly for Graph Representation? , url =

    work page

  15. [15]

    2025 , eprint=

    YOLOv13: Real-Time Object Detection with Hypergraph-Enhanced Adaptive Visual Perception , author=. 2025 , eprint=

    work page 2025

  16. [16]

    SK-GCN: Modeling Syntax and Knowledge via Graph Convolutional Network for aspect-level sentiment classification , journal =

    Jie Zhou and Jimmy Xiangji Huang and Qinmin Vivian Hu and Liang He , keywords =. SK-GCN: Modeling Syntax and Knowledge via Graph Convolutional Network for aspect-level sentiment classification , journal =. 2020 , issn =. doi:https://doi.org/10.1016/j.knosys.2020.106292 , url =

    work page doi:10.1016/j.knosys.2020.106292 2020

  17. [17]

    Inducing Target-Specific Latent Structures for Aspect Sentiment Classification

    Chen, Chenhua and Teng, Zhiyang and Zhang, Yue. Inducing Target-Specific Latent Structures for Aspect Sentiment Classification. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP). 2020. doi:10.18653/v1/2020.emnlp-main.451

    work page doi:10.18653/v1/2020.emnlp-main.451 2020

  18. [18]

    Relational Graph Attention Network for Aspect-based Sentiment Analysis

    Wang, Kai and Shen, Weizhou and Yang, Yunyi and Quan, Xiaojun and Wang, Rui. Relational Graph Attention Network for Aspect-based Sentiment Analysis. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020. doi:10.18653/v1/2020.acl-main.295

    work page doi:10.18653/v1/2020.acl-main.295 2020

  19. [19]

    Dependency Graph Enhanced Dual-transformer Structure for Aspect-based Sentiment Classification

    Tang, Hao and Ji, Donghong and Li, Chenliang and Zhou, Qiji. Dependency Graph Enhanced Dual-transformer Structure for Aspect-based Sentiment Classification. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020. doi:10.18653/v1/2020.acl-main.588

    work page doi:10.18653/v1/2020.acl-main.588 2020

  20. [20]

    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

  21. [21]

    Aspect-based Sentiment Analysis with Type-aware Graph Convolutional Networks and Layer Ensemble

    Tian, Yuanhe and Chen, Guimin and Song, Yan. Aspect-based Sentiment Analysis with Type-aware Graph Convolutional Networks and Layer Ensemble. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021. doi:10.18653/v1/2021.naacl-main.231

    work page doi:10.18653/v1/2021.naacl-main.231 2021

  22. [22]

    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

  23. [23]

    Discrete Opinion Tree Induction for Aspect-based Sentiment Analysis

    Chen, Chenhua and Teng, Zhiyang and Wang, Zhongqing and Zhang, Yue. Discrete Opinion Tree Induction for Aspect-based Sentiment Analysis. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022. doi:10.18653/v1/2022.acl-long.145

    work page doi:10.18653/v1/2022.acl-long.145 2022

  24. [24]

    The Journal of Supercomputing , volume =

    Yu, Bengong and Zhang, Shuwen , title =. The Journal of Supercomputing , volume =. 2023 , doi =

    work page 2023

  25. [25]

    and Wen, Lijie

    Ma, Fukun and Hu, Xuming and Liu, Aiwei and Yang, Yawen and Li, Shuang and Yu, Philip S. and Wen, Lijie. AMR -based Network for Aspect-based Sentiment Analysis. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023. doi:10.18653/v1/2023.acl-long.19

    work page doi:10.18653/v1/2023.acl-long.19 2023

  26. [26]

    S ^2 GSL : Incorporating Segment to Syntactic Enhanced Graph Structure Learning for Aspect-based Sentiment Analysis

    Chen, Bingfeng and Ouyang, Qihan and Luo, Yongqi and Xu, Boyan and Cai, Ruichu and Hao, Zhifeng. S ^2 GSL : Incorporating Segment to Syntactic Enhanced Graph Structure Learning for Aspect-based Sentiment Analysis. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024. doi:10.18653/v1/2024.acl...

    work page doi:10.18653/v1/2024.acl-long.721 2024

  27. [27]

    Aspect-based sentiment analysis with semantic and syntactic enhanced multi-layer fusion model , journal =

    Song Jin and Qing He and Yuji Wang and Nisuo Du and Wenjing Lei , keywords =. Aspect-based sentiment analysis with semantic and syntactic enhanced multi-layer fusion model , journal =. 2025 , issn =. doi:https://doi.org/10.1016/j.engappai.2025.111654 , url =

    work page doi:10.1016/j.engappai.2025.111654 2025

  28. [28]

    Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =

    Zhang, Chen and Li, Qiuchi and Song, Dawei , title =. Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =. 2019 , isbn =. doi:10.1145/3331184.3331351 , abstract =

    work page doi:10.1145/3331184.3331351 2019

  29. [29]

    Does syntax matter? A strong baseline for Aspect-based Sentiment Analysis with R o BERT a

    Dai, Junqi and Yan, Hang and Sun, Tianxiang and Liu, Pengfei and Qiu, Xipeng. Does syntax matter? A strong baseline for Aspect-based Sentiment Analysis with R o BERT a. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021. doi:10.18653/v1/2021.naacl-main.146

    work page doi:10.18653/v1/2021.naacl-main.146 2021

  30. [30]

    Journal of Big Data , volume =

    Feng, Ao and Cai, Jiazhi and Gao, Zhengjie and Li, Xiaojie , title =. Journal of Big Data , volume =. 2023 , doi =

    work page 2023

  31. [31]

    Eliminating Sentiment Bias for Aspect-Level Sentiment Classification with Unsupervised Opinion Extraction

    Wang, Bo and Shen, Tao and Long, Guodong and Zhou, Tianyi and Chang, Yi. Eliminating Sentiment Bias for Aspect-Level Sentiment Classification with Unsupervised Opinion Extraction. Findings of the Association for Computational Linguistics: EMNLP 2021. 2021. doi:10.18653/v1/2021.findings-emnlp.258

    work page doi:10.18653/v1/2021.findings-emnlp.258 2021

  32. [32]

    Local aggressive and physically realizable adversarial attacks on 3d point cloud

    Zhongquan Jian and Jiajian Li and Qingqiang Wu and Junfeng Yao , keywords =. Retrieval Contrastive Learning for Aspect-Level Sentiment Classification , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.ipm.2023.103539 , url =

    work page doi:10.1016/j.ipm.2023.103539 2024

  33. [33]

    AGCL : Aspect Graph Construction and Learning for Aspect-level Sentiment Classification

    Jian, Zhongquan and Wu, Daihang and Wang, Shaopan and Wang, Yancheng and Yao, Junfeng and Wang, Meihong and Wu, Qingqiang. AGCL : Aspect Graph Construction and Learning for Aspect-level Sentiment Classification. Proceedings of the 31st International Conference on Computational Linguistics. 2025

    work page 2025

  34. [34]

    Learning Implicit Sentiment in Aspect-based Sentiment Analysis with Supervised Contrastive Pre-Training

    Li, Zhengyan and Zou, Yicheng and Zhang, Chong and Zhang, Qi and Wei, Zhongyu. Learning Implicit Sentiment in Aspect-based Sentiment Analysis with Supervised Contrastive Pre-Training. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. 2021. doi:10.18653/v1/2021.emnlp-main.22

    work page doi:10.18653/v1/2021.emnlp-main.22 2021

  35. [35]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    Aspect-Based Sentiment Analysis with Explicit Sentiment Augmentations , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2024 , month=. doi:10.1609/aaai.v38i17.29849 , abstractNote=

    work page doi:10.1609/aaai.v38i17.29849 2024

  36. [36]

    Reasoning Implicit Sentiment with Chain-of-Thought Prompting

    Fei, Hao and Li, Bobo and Liu, Qian and Bing, Lidong and Li, Fei and Chua, Tat-Seng. Reasoning Implicit Sentiment with Chain-of-Thought Prompting. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers). 2023. doi:10.18653/v1/2023.acl-short.101

    work page doi:10.18653/v1/2023.acl-short.101 2023

  37. [37]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    Counterfactual-Enhanced Information Bottleneck for Aspect-Based Sentiment Analysis , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2024 , month=. doi:10.1609/aaai.v38i16.29726 , abstractNote=

    work page doi:10.1609/aaai.v38i16.29726 2024

  38. [38]

    Modeling Aspect Sentiment Coherency via Local Sentiment Aggregation

    Yang, Heng and Li, Ke. Modeling Aspect Sentiment Coherency via Local Sentiment Aggregation. Findings of the Association for Computational Linguistics: EACL 2024. 2024. doi:10.18653/v1/2024.findings-eacl.13

    work page doi:10.18653/v1/2024.findings-eacl.13 2024

  39. [39]

    Applied Sciences , VOLUME =

    Zeng, Biqing and Yang, Heng and Xu, Ruyang and Zhou, Wu and Han, Xuli , TITLE =. Applied Sciences , VOLUME =. 2019 , NUMBER =

    work page 2019

  40. [40]

    Tasty Burgers, Soggy Fries: Probing Aspect Robustness in Aspect-Based Sentiment Analysis

    Xing, Xiaoyu and Jin, Zhijing and Jin, Di and Wang, Bingning and Zhang, Qi and Huang, Xuanjing. Tasty Burgers, Soggy Fries: Probing Aspect Robustness in Aspect-Based Sentiment Analysis. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP). 2020. doi:10.18653/v1/2020.emnlp-main.292

    work page doi:10.18653/v1/2020.emnlp-main.292 2020

  41. [41]

    Reproducibility, Replicability and Beyond: Assessing Production Readiness of Aspect Based Sentiment Analysis in the Wild

    Mukherjee, Rajdeep and Shetty, Shreyas and Chattopadhyay, Subrata and Maji, Subhadeep and Datta, Samik and Goyal, Pawan. Reproducibility, Replicability and Beyond: Assessing Production Readiness of Aspect Based Sentiment Analysis in the Wild. Advances in Information Retrieval. 2021. doi:10.1007/978-3-030-72240-1_7 , url =

    work page doi:10.1007/978-3-030-72240-1_7 2021

  42. [42]

    Proceedings of the 32nd ACM International Conference on Information and Knowledge Management , pages =

    Yang, Heng and Zhang, Chen and Li, Ke , title =. Proceedings of the 32nd ACM International Conference on Information and Knowledge Management , pages =. 2023 , isbn =. doi:10.1145/3583780.3614752 , abstract =

    work page doi:10.1145/3583780.3614752 2023

  43. [43]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    Hypergraph Neural Networks , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2019 , month=. doi:10.1609/aaai.v33i01.33013558 , abstractNote=

    work page doi:10.1609/aaai.v33i01.33013558 2019

  44. [44]

    DAGF: A dual GCN and auxiliary graph fusion based model for aspect-based sentiment analysis , journal =

    Jie Ji and Wenlong Zhu and Chengle Hou and Qiaoyan Song and YuKun Ma and Youruo Wang , keywords =. DAGF: A dual GCN and auxiliary graph fusion based model for aspect-based sentiment analysis , journal =. 2026 , issn =. doi:https://doi.org/10.1016/j.asoc.2026.115040 , url =

    work page doi:10.1016/j.asoc.2026.115040 2026

  45. [45]

    DeBERTa: Decoding-enhanced BERT with Disentangled Attention

    Pengcheng He and Xiaodong Liu and Jianfeng Gao and Weizhu Chen , title =. CoRR , volume =. 2020 , url =. 2006.03654 , timestamp =

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  46. [46]

    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

  47. [47]

    Adaptive Recursive Neural Network for Target-dependent T witter Sentiment Classification

    Dong, Li and Wei, Furu and Tan, Chuanqi and Tang, Duyu and Zhou, Ming and Xu, Ke. Adaptive Recursive Neural Network for Target-dependent T witter Sentiment Classification. Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers). 2014. doi:10.3115/v1/P14-2009

    work page doi:10.3115/v1/p14-2009 2014

  48. [48]

    S em E val-2015 Task 12: Aspect Based Sentiment Analysis

    Pontiki, Maria and Galanis, Dimitris and Papageorgiou, Haris and Manandhar, Suresh and Androutsopoulos, Ion. S em E val-2015 Task 12: Aspect Based Sentiment Analysis. Proceedings of the 9th International Workshop on Semantic Evaluation ( S em E val 2015). 2015. doi:10.18653/v1/S15-2082

    work page doi:10.18653/v1/s15-2082 2015

  49. [49]

    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

  50. [50]

    A Challenge Dataset and Effective Models for Aspect-Based Sentiment Analysis

    Jiang, Qingnan and Chen, Lei and Xu, Ruifeng and Ao, Xiang and Yang, Min. A Challenge Dataset and Effective Models for Aspect-Based Sentiment Analysis. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019. doi:10.18653/v1/D19-1654

    work page doi:10.18653/v1/d19-1654 2019