Pith. sign in

REVIEW 6 major objections 5 minor 46 references

AdaptiSent: Context-Aware Adaptive Attention for Multimodal Aspect-Based Sentiment Analysis

T0 review · 6 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read AdaptiSent claims per-aspect adaptive text-image weighting achieves new state-of-the-art F1 on Twitter-15 and Twitter-17.

desk verdict Reasonable MABSA engineering, but the SOTA claim is contradicted by the paper's own Table 4; revise rather than desk-reject. read the letter →

arxiv 2507.12695 v1 pith:HWAZA2CU submitted 2025-07-17 cs.CL

S M Rafiuddin , Sadia Kamal , Mohammed Rakib , Arunkumar Bagavathi , Atriya Sen This is my paper · ORCID
classification cs.CL
keywords multimodalaspect-basedsentimentanalysisadaptivecross-modalattentionaspecttermextractionclassificationmodalityalignmentTwitter-15Twitter-17dataaugmentation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

AdaptiSent is a framework for multimodal aspect-based sentiment analysis, the task of finding which specific targets a tweet praises or criticizes and reading the image alongside the text. The paper's claim is that letting the model decide, for each aspect term, how much to trust the text versus the image, rather than fusing the two modalities in a fixed way, improves both aspect extraction and sentiment prediction. On the two standard Twitter benchmarks, the full model reports F1 of 71.89 on Twitter-15 and 71.62 on Twitter-17, above every baseline listed in the comparison. The paper also isolates which design choices carry the gain: per-aspect balancing coefficients, aspect-aware captions, the alignment regularizer, and context masking all matter, while LLM data augmentation adds little. The paper presents this as evidence that per-aspect adaptive weighting offers a simple route to better multimodal sentiment understanding.

What carries the argument

The load-bearing object is the context-adaptive cross-modal attention mechanism. It computes per-token importance $S(t_i) = \gamma R_{\text{ling}}(t_i) + (1-\gamma) R_{\text{vis}}(t_i)$, where $R_{\text{ling}}$ comes from a learned combination of syntactic and NER embeddings and $R_{\text{vis}}$ from attention between text tokens and both image patches and aspect-aware captions, then adds $\beta S(t_i)$ to the scaled dot-product $\frac{QK^\top}{\sqrt{d_k}}$ before the softmax. This additive bias lets the model up- or down-weight tokens that are linguistically or visually salient for the current aspect. Three supporting mechanisms carry the rest of the argument: an adaptive masking threshold $\theta = \mu_S + \alpha_m \sigma_S$ that blanks out irrelevant tokens per sentence, a trainable aspect-specific balancing coefficient that weights text versus image embeddings, and a squared-Euclidean modality alignment regularizer that ties the two modalities to a shared embedding space.

What would settle it

Re-run the strongest prior model DQPSA and AdaptiSent under identical training conditions, splits, seeds, and decoding rules on Twitter-15 and Twitter-17; if DQPSA's F1 matches or exceeds 71.89 on Twitter-15 within one standard deviation, the paper's central 'surpasses existing models' claim fails as stated.

Watch

Extended reading notes

Core claim

The central claim is that AdaptiSent achieves a new state of the art in multimodal aspect-based sentiment analysis on Twitter-15 and Twitter-17. It does so by treating the text-image relationship as something to be learned separately for every aspect rather than applied uniformly. For each candidate aspect, the model computes a linguistic importance score from dependency, part-of-speech, and named-entity cues, and a visual-to-text relevance score from attention between tokens, image patches, and aspect-aware captions. These scores are combined into a single importance signal that biases the cross-modal attention, and a trainable coefficient decides how much the final representation leans on text versus image for that aspect. A squared-distance regularizer pulls the projected text and image embeddings into a shared space. The paper reports the strongest F1 scores in its comparison table and shows, via ablation, that the per-aspect balancing coefficient is the largest single contributor to the gain.

Load-bearing premise

The load-bearing premise is that the baseline numbers in Table 4 were produced under the same evaluation protocol as AdaptiSent's, since the paper reports no re-runs, no error bars for prior systems, and no significance tests.

Editorial extensions

If this is right

  • On the Twitter-15 and Twitter-17 benchmarks, the full model reaches F1 of 71.89 and 71.62, the highest figures reported in the comparison.
  • Removing the aspect-specific balancing coefficients hurts most in the ablation, so the paper's analysis says per-aspect adaptive weighting is the main source of the gain.
  • Aspect-aware captions and the modality-alignment regularizer each give substantial secondary gains, while LLM-based data augmentation contributes little.
  • In the paper's case studies, the model agrees with ground truth on all four examples, including cases where three comparison systems made errors.
  • Because the attention modulation adds only token-level importance scores to the standard attention computation, the claimed gains come without adding attention heads or layers.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural check, given the reported numbers, is a significance test against the strongest baseline DQPSA, whose Twitter-15 F1 is essentially the same as AdaptiSent's in the table.
  • The adaptive masking rule $\theta = \mu_S + \alpha_m \sigma_S$ is a general instance-dependent token-selection mechanism that could transfer to other tasks, such as rationale extraction or keyphrase generation.
  • Because visual-guided augmentation depends on external LLMs, a reproduction with open-weight models only would test how much of the gain is tied to the specific generators.
  • The squared-Euclidean alignment regularizer could be stress-tested on deliberately misaligned image-text pairs, a setting the paper names as future work.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

6 major / 5 minor

Summary. The paper introduces AdaptiSent, a framework for multimodal aspect-based sentiment analysis (MABSA) that combines a RoBERTa text encoder, a ViT image encoder, dynamic importance scoring, adaptive masking, aspect-specific balancing coefficients, LLM-based data augmentation, and a modality-alignment regularizer. The method is evaluated on Twitter-15 and Twitter-17, and the authors claim that AdaptiSent surpasses existing models in precision, recall, and F1 and sets a new state-of-the-art. The manuscript includes ablations, hyperparameter sensitivity plots, and three-seed standard deviations for the proposed model, but it does not provide public code or data, baseline error bars, or significance tests.

Significance. If the performance claims held, the framework would offer a plausible set of components for MABSA, particularly the adaptive importance scoring, the aspect-specific balancing coefficients, and the modality-alignment regularizer. The paper reports ablations and hyperparameter sensitivity, which are useful for understanding the contribution of each component. However, the central empirical claim is not supported by the evidence as presented: Table 4 shows a tie on Twitter-15 F1 and lower precision than DQPSA, and the differences on Twitter-17 are small relative to the reported standard deviations. The absence of baseline error bars, significance tests, and public code or data further limits verification. The design is coherent, but the claimed state-of-the-art significance is not established.

major comments (6)
  1. [Abstract; §5.1–5.2, Table 4] The abstract and Section 5.1 claim that AdaptiSent "surpasses existing models in precision, recall, and F1 score" and "sets a new state-of-the-art," but Table 4 contradicts this claim on Twitter-15: DQPSA has F1 71.9, equal to the reported 71.9, and precision 71.7, higher than AdaptiSent's 70.9; only recall favors AdaptiSent. Please either restrict the claim to the specific metrics and datasets where the improvement actually holds, or provide additional evidence, such as paired significance tests over multiple seeds, that supports the broader claim.
  2. [Table 4] The paper reports standard deviations for AdaptiSent only; no baseline error bars, significance tests, or a statement of the baseline evaluation protocol are provided. The largest F1 advantage over the best non-LLM baseline on Twitter-17 is 0.6 points over TMFN, which is comparable to AdaptiSent's own reported standard deviations of 0.18–0.52. Without knowing whether the baselines were run under the same preprocessing, splitting, and evaluation script, and without paired significance tests, the observed differences are not established as real improvements. Please re-run the baselines under the same environment or clearly state that the numbers are quoted from the original papers, and report significance tests or confidence intervals.
  3. [§3.5.1, Eq. (18)] The joint training objective is under-specified. Equation (18) displays a single token-level cross-entropy term plus the alignment regularizer, but the task requires both aspect term extraction and sentiment classification. There is no explicit loss term for sentiment polarity or for the aspect-extraction objective, and the token-specific weight w_i is not defined in terms of the model components. Please specify the complete multi-task loss, including how the aspect-term extraction and sentiment-classification losses are computed and combined.
  4. [§3.3.1, Eqs. (5)–(6)] The adaptive masking rule appears to mask the most important tokens: Eq. (6) replaces tokens with S(t_i) > θ by [MASK]. Since S is the combined importance score used elsewhere to focus on relevant cues, this contradicts the stated motivation and is likely to harm aspect extraction. Please clarify whether the inequality should be reversed, and include an ablation comparing both masking directions.
  5. [§3.4.1, Eq. (8)] The data-augmentation step conditions on A_ext, the candidate aspects produced by the model itself during training. If those predictions are noisy, the augmented examples can reinforce extraction errors, and the paper does not report any filtering or verification of the LLM-generated text. Section 5.2 shows that removing augmentation has only a small effect (–1.51 and –0.69 F1 points on Twitter-15 and Twitter-17), so the component's benefit is marginal; please analyze its impact on extraction errors or reconsider whether it should be listed as a key contribution.
  6. [§3.2, Fig. 1] The aspect-aware captions C0 are used throughout the model, including in Eq. (3) and Eq. (7), but the manuscript never specifies how C0 is generated from the input image and candidate aspects. The figure attributes caption generation to CLIP, but no prompt, decoding procedure, or curation/filtering step is described. Please provide a precise description of the captioning component so that the method is reproducible.
minor comments (5)
  1. [Title page] The first page contains the typo "ACCEPETD" where "ACCEPTED" is intended.
  2. [§5.3, Table 6] The model name is spelled inconsistently: "DPQSA" appears in Section 5.3 and Table 6, while "DQPSA" appears in Table 4 and in the baseline descriptions. Please use one spelling throughout.
  3. [Table 6] Table 6 uses checkmark and cross symbols that are not explained in the caption. Please define what ✓ and ✗ denote and how the reader should interpret the cell-level entries.
  4. [§4.2] The experimental setup does not state the number of training epochs, warmup steps, or early-stopping criterion. Adding these details would improve reproducibility.
  5. [Eq. (9)] The coherence score in Eq. (9) is computed between the augmented text embedding and the visual embedding, but the text does not explain how this coherence value is used in the training loss or in filtering augmented examples. Please specify its role.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the model is evaluated on external benchmarks and its components are standard supervised learning pieces, though the LLM data augmentation is a mild self-training loop that is not load-bearing.

full rationale

AdaptiSent's design is a conventional supervised MABSA pipeline: text and image embeddings (RoBERTa, ViT, CLIP) are fixed pretrained features; importance scores, adaptive masking, modality weighting, and the alignment regularizer are all computed from the model's own inputs and trainable parameters, not from the target labels. The final predictions are compared against held-out test sets (Twitter-15, Twitter-17) with ground-truth aspect–sentiment labels, so the central F1/recall/precision claims are externally grounded rather than derived from the model's assumptions. The only self-referential component is the visual-guided data augmentation (Eq. 8): the model's own extracted aspects Aext condition LLM-generated augmented text that is then added to the training set. This is a form of self-training and could in principle create a self-reinforcing loop, but the ablation (§5.2) shows removing data augmentation changes F1 by only –1.51 and –0.69 points, so it is not load-bearing for the main result. There are no self-citations to prior work by the same authors, no imported uniqueness theorem, and no fitted parameter renamed as a prediction. The paper's overclaim of state-of-the-art (Table 4 shows a tie with DQPSA on Twitter-15 F1 and lower precision) is a correctness/comparability issue, not circularity, and is outside the scope of this pass. The derivation chain is therefore self-contained at the level of the actual reported experiments.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The model introduces several trainable parameters and two validation-tuned hyperparameters. No new physical or conceptual entities are postulated. The main assumptions are that pretrained encoders and LLM augmentation help, and that the baseline numbers are comparable.

free parameters (2)
  • gamma = 0.3
    Hyperparameter balancing linguistic vs visual importance (Eq. 10), tuned on validation data.
  • lambda = 0.1
    Regularization strength for modality alignment (Eq. 17), tuned on validation.
assumptions (4)
  • domain assumption Pretrained RoBERTa and ViT embeddings transfer to the MABSA task
    Section 4.2 initializes models with pretrained weights; the method assumes these features are useful for aspect sentiment.
  • domain assumption Aspect-aware captions from CLIP provide complementary semantics
    Section 3.2 and Figure 1 use CLIP to generate captions without describing the generation procedure or validating its contribution.
  • ad hoc to paper The adaptive masking rule (Eq. 5-6) that masks high-importance tokens improves extraction
    The threshold formula is introduced without theoretical justification; it is an empirical design choice.
  • domain assumption The compared baselines' results are directly applicable to this evaluation
    Section 5.1 lists numbers from prior papers without re-running baselines under identical conditions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AdaptiSent: Context-Aware Adaptive Attention for Multimodal Aspect-Based Sentiment Analysis." pith.science (2026). https://pith.science/paper/HWAZA2CU

@misc{pith2026250712695,
  author       = {Pith},
  title        = {Pith review of: AdaptiSent: Context-Aware Adaptive Attention for Multimodal Aspect-Based Sentiment Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HWAZA2CU}},
  note         = {Machine review of arXiv:2507.12695}
}
read the original abstract

We introduce AdaptiSent, a new framework for Multimodal Aspect-Based Sentiment Analysis (MABSA) that uses adaptive cross-modal attention mechanisms to improve sentiment classification and aspect term extraction from both text and images. Our model integrates dynamic modality weighting and context-adaptive attention, enhancing the extraction of sentiment and aspect-related information by focusing on how textual cues and visual context interact. We tested our approach against several baselines, including traditional text-based models and other multimodal methods. Results from standard Twitter datasets show that AdaptiSent surpasses existing models in precision, recall, and F1 score, and is particularly effective in identifying nuanced inter-modal relationships that are crucial for accurate sentiment and aspect term extraction. This effectiveness comes from the model's ability to adjust its focus dynamically based on the context's relevance, improving the depth and accuracy of sentiment analysis across various multimodal data sets. AdaptiSent sets a new standard for MABSA, significantly outperforming current methods, especially in understanding complex multimodal information.

Figures

Figures reproduced from arXiv: 2507.12695 by the authors.

Figure 1
Figure 1. Overview of the AdaptiSent framework for MABSA. Given a tweet and its paired image, an LLM augments the input with aspect terms. Linguistic features (dependency, POS, NER) guide adaptive masking, and the masked text is encoded by RoBERTa. Simultaneously, CLIP [44] generates aspect-aware captions and ViT extracts patch-level visual features. A visual-to-text relevance module assigns importance scores, fused via cross… view at source ↗
Figure 2
Figure 2. Hyperparameter sensitivity: (a) variation with [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 40 canonical work pages

  1. [1]

    and Na, J

    Yang, L. and Na, J. C. and Yu, J.. Cross-modal multitask transformer for end-to-end multimodal aspect-based sentiment analysis. Information Processing & Management, 59(5): 103038, 2022

    work page 2022

  2. [2]

    and Wu, X

    Xiao, L. and Wu, X. and Xu, J. and Li, W. and Jin, C. and He, L.. Atlantis: Aesthetic-oriented multiple granularities fusion network for joint multimodal aspect-based sentiment analysis. Information Fusion, 106: 102304, 2024

    work page 2024

  3. [3]

    and Sun, H

    Zhu, L. and Sun, H. and Gao, Q. and Yi, T. and He, L.. Joint multimodal aspect sentiment analysis with aspect enhancement and syntactic adaptive learning. In Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence (IJCAI), 2015

    work page 2015

  4. [4]

    and Zhao, Y

    Yang, H. and Zhao, Y . and Qin, B.. Face-sensitive image-to-emotional-text cross-modal translation for multimodal aspect-based sentiment analysis. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 3324–3335, 2022

    work page 2022

  5. [5]

    and Lin, M

    Feng, J. and Lin, M. and Shang, L. and Gao, X.. Autonomous aspect-image instruction a2II: Q-Former guided multimodal sentiment classification. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pp. 1996–2005, 2024

    work page 2024

  6. [6]

    and He, Y

    Wang, D. and He, Y . and Liang, X. and Tian, Y . and Li, S. and Zhao, L.. TMFN: A target-oriented multi-grained fusion network for end-to-end aspect-based multimodal sentiment analysis. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pp. 16187–16197, 2024

    work page 2024

  7. [7]

    A vision-language pre-training model based on cross attention for multimodal aspect-based sentiment analysis

    Hu, H.. A vision-language pre-training model based on cross attention for multimodal aspect-based sentiment analysis. In 2024 5th International Conference on Computer Vision, Image and Deep Learning (CVIDL), pp. 370–375, 2024

    work page 2024

  8. [8]

    and Chen, J

    Fan, H. and Chen, J.. Position perceptive multi-hop fusion network for multimodal aspect-based sentiment analysis. IEEE Access, 2024

    work page 2024

Show all 46 references
  1. [9]

    Vargas, D. S. and Pessutto, L. R. C. and Moreira, V . P. and de Melo, T. and Da Silva, A. S.. EurOpi: Multilingual aspect-based sentiment analysis enabled by a knowledge base. In 2022 IEEE/WIC/ACM International Joint Conference on Web Intelligence and Intelligent Agent Technol...

    2022

  2. [10]

    and Feng, S

    Yang, X. and Feng, S. and Wang, D. and Sun, Q. and Wu, W. and Zhang, Y . and Hong, P. and Poria, S.. Few-shot joint multimodal aspect-sentiment analysis based on generative multimodal prompt. In Findings of the Association for Computational Linguistics: ACL 2023, pp. 11575–11589, 2023

    2023

  3. [11]

    and Cai, Y

    Xiang, Y . and Cai, Y . and Guo, J.. MSFNet: Modality smoothing fusion network for multimodal aspect-based sentiment analysis. Frontiers in Physics, 11: 1187503, 2023

    2023

  4. [12]

    and Sharma, A

    Chauhan, A. and Sharma, A. and Mohana, R.. A transformer model for end-to-end image and text aspect-based sentiment analysis. In 2023 Seventh International Conference on Image Information Processing (ICIIP), pp. 277–282, 2023

    2023

  5. [13]

    and Su, Q

    Xu, Z. and Su, Q. and Xiao, J.. Multimodal aspect-based sentiment classification with knowledge-injected transformer. In 2023 IEEE International Conference on Multimedia and Expo (ICME), pp. 1379–1384, 2023

    2023

  6. [14]

    and Feng, H

    Zhou, Z. and Feng, H. and Qiao, B. and Wu, G. and Han, D.. Syntax-aware hybrid prompt model for few-shot multi-modal sentiment analysis. arXiv preprint arXiv:2306.01312, 2023

    2023 arXiv

  7. [15]

    and Li, C

    Zhao, F. and Li, C. and Wu, Z. and Ouyang, Y . and Zhang, J. and Dai, X.. M2DF: Multi-grained multi-curriculum denoising framework for multimodal aspect-based sentiment analysis. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP),...

    2023

  8. [16]

    and Li, Z

    Peng, T. and Li, Z. and Wang, P. and Zhang, L. and Zhao, H.. A novel energy based model mechanism for multi-modal aspect-based sentiment analysis. In Proceedings of the AAAI Conference on Artificial Intelligence, pp. 18869–18878, 2024

    2024

  9. [17]

    and Guo, J

    Wang, Z. and Guo, J.. Self-adaptive attention fusion for multimodal aspect-based sentiment analysis. Mathematical Biosciences and Engineering, 21(1): 1305–1320, 2024

    2024

  10. [18]

    RoBERTa: A robustly optimized BERT

    Liu, Y .. RoBERTa: A robustly optimized BERT. arXiv preprint arXiv:1907.11692, 2019

    1907 arXiv

  11. [19]

    Decoupled weight decay regularization

    Loshchilov, I.. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017. 10 AdaptiSent: Context-Aware Adaptive Attention for Multimodal ABSA ASONAM 2025

    2017 arXiv

  12. [20]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Dosovitskiy, A.. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020

    2010 arXiv

  13. [21]

    and Peng, Y

    Hu, M. and Peng, Y . and Huang, Z. and Li, D. and Lv, Y .. Open-domain targeted sentiment analysis via span-based extraction and classification. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 537–546, 2019

    2019

  14. [22]

    and Tian, Y

    Chen, G. and Tian, Y . and Song, Y .. Joint aspect extraction and sentiment analysis with directional graph convolutional networks. In Proceedings of the 28th International Conference on Computational Linguistics, pp. 272–279, 2020

    2020

  15. [23]

    BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension

    Lewis, M.. BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. arXiv preprint arXiv:1910.13461, 2019

    1910 arXiv

  16. [24]

    and Jiang, J

    Yu, J. and Jiang, J. and Yang, L. and Xia, R.. Improving multimodal named entity recognition via entity span detection with unified multimodal transformer. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pp. 3342–3352, 2020

    2020

  17. [25]

    and Zheng, C

    Wu, Z. and Zheng, C. and Cai, Y . and Chen, J. and Leung, H. and Li, Q.. Multimodal representation with embedded visual guiding objects for named entity recognition in social media posts. In Proceedings of the 28th ACM International Conference on Multimedia, pp. 1038–1046, 2020

    2020

  18. [26]

    and Zhang, D

    Ju, X. and Zhang, D. and Xiao, R. and Li, J. and Li, S. and Zhang, M. and Zhou, G.. Joint multi-modal aspect- sentiment analysis with auxiliary cross-modal relation detection. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp...

    2021

  19. [27]

    and Yu, J

    Ling, Y . and Yu, J. and Xia, R.. Vision-language pre-training for multimodal aspect-based sentiment analysis. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (V olume 1: Long Papers), pp. 2149–2159, 2022

    2022

  20. [28]

    and Wang, J

    Yu, Z. and Wang, J. and Yu, L. C. and Zhang, X.. Dual-encoder transformers with cross-modal alignment for multimodal aspect-based sentiment analysis. In Proceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th ...

    2022

  21. [29]

    and Martin, L

    Touvron, H. and Martin, L. and Stone, K. and Albert, P. and Almahairi, A. and Babaei, Y . and Bashlykov, N. and Batra, S. and Bhargava, P. and Bhosale, S. and others. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023

    2023 arXiv

  22. [30]

    ChatGPT: A large language model

    OpenAI. ChatGPT: A large language model. 2023

    2023

  23. [31]

    and Jiang, J

    Yu, J. and Jiang, J. Adapting BERT. In Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence (IJCAI), pp. 5408–5414, 2019

    2019

  24. [32]

    and Jiang, J

    Yu, J. and Jiang, J. and Xia, R.. Entity-sensitive attention and fusion network for entity-level multimodal sentiment classification. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 28: 429–439, 2019

    2019

  25. [33]

    and Fu, Y

    Khan, Z. and Fu, Y .. Exploiting BERT. In Proceedings of the 29th ACM International Conference on Multimedia, pp. 3034–3042, 2021

    2021

  26. [34]

    and Qian, Y

    Du, Z. and Qian, Y . and Liu, X. and Ding, M. and Qiu, J. and Yang, Z. and Tang, J.. GLM: General language model pretraining with autoregressive blank infilling. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (V olume 1: Long Papers)...

    2022

  27. [35]

    and Wang, Z

    Zhang, Z. and Wang, Z. and Li, X. and Liu, N. and Guo, B. and Yu, Z.. ModalNet: An aspect-level sentiment classification model by exploring multimodal data with fusion discriminant attentional network. World Wide Web, 24: 1957–1974, 2021

    1957

  28. [36]

    and Liu, Z

    Wang, J. and Liu, Z. and Sheng, V . and Song, Y . and Qiu, C.. SaliencyBERT: Recurrent attention network for target-oriented multimodal sentiment classification. In Pattern Recognition and Computer Vision: 4th Chinese Conference, PRCV 2021, Beijing, China, October 29–November ...

    2021

  29. [37]

    and Huang, M

    Wang, Y . and Huang, M. and Zhu, X. and Zhao, L.. Attention-based LSTM for aspect-level sentiment classification. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 606–615, 2016

    2016

  30. [38]

    and Qin, B

    Tang, D. and Qin, B. and Liu, T.. Aspect level sentiment classification with deep memory network. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 214–224, 2016

    2016

  31. [39]

    and Sun, Z

    Chen, P. and Sun, Z. and Bing, L. and Yang, W.. Recurrent attention network on memory for aspect sentiment analysis. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 452–461, 2017. 11 AdaptiSent: Context-Aware Adaptive Atte...

    2017

  32. [40]

    and Feng, Y

    Fan, F. and Feng, Y . and Zhao, D.. Multi-grained attention network for aspect-level sentiment classification. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 3433–3442, 2018

    2018

  33. [41]

    BERT: Pre-training of deep bidirectional transformers for language understanding

    Devlin, J.. BERT: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018

    2018 arXiv

  34. [42]

    and Zhou, E

    Xiao, L. and Zhou, E. and Wu, X. and Yang, S. and Ma, T. and He, L.. Adaptive multi-feature extraction graph convolutional networks for multimodal target sentiment analysis. In 2022 IEEE International Conference on Multimedia and Expo (ICME), pp. 1–6, 2022

    2022

  35. [43]

    and Wu, X

    Xiao, L. and Wu, X. and Yang, S. and Xu, J. and Zhou, J. and He, L.. Cross-modal fine-grained alignment and fusion network for multimodal aspect-based sentiment analysis. Information Processing & Management, 60(6): 103508, 2023

    2023

  36. [44]

    and Kim, J

    Radford, A. and Kim, J. W. and Hallacy, C. and Ramesh, A. and Goh, G. and Agarwal, S. and Sastry, G. and Askell, A. and Mishkin, P. and Clark, J. and others. Learning transferable visual models from natural language supervision. In International Conference on Machine Learning ...

    2021

  37. [45]

    Attention is all you need

    Vaswani, A.. Attention is all you need. In Advances in Neural Information Processing Systems, 2017

    2017

  38. [46]

    and Guo, W

    Zhou, R. and Guo, W. and Liu, X. and Yu, S. and Zhang, Y . and Yuan, X.. AoM: Detecting aspect-oriented information for multimodal aspect-based sentiment analysis. In Findings of the Association for Computational Linguistics: ACL 2023, pp. 8184–8196, 2023. 12

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.