Modality Modulation with Adaptive Fusion for Multimodal Sentiment Analysis
Bibliographic record
Abstract
Multimodal sentiment analysis aims at extracting effective information from different modalities such as text, audio, and visual to infer the speaker’s sentiment state. Due to the high heterogeneity across modalities, most existing approaches decouple modalities into specific and invariant features, which can capture effective cross-modal representations to some extent. However, in multimodal tasks, different modalities exhibit varying strengths and weaknesses, with strong modalities dominating the overall optimization direction of the network, leading to under-optimization of weak modalities. To address the modality imbalance issue, we propose a Modality Modulation Adaptive Fusion Network (MMAFNet) to optimize the learning of valuable information from each modality. Specifically, for modality-specific features, we design a specific feature gradient modulation strategy to stimulate the weak modalities learning and adaptively modulate the corresponding gradients by measuring different importance to better optimize each modality. For modality-invariant features, in terms of the distance between different modalities, we propose an invariant feature parameter reset strategy that prevents overfitting irrelevant information while enhancing the feature extraction capability of weaker modalities. Finally, we incorporate an adaptive fusion module to combine modality-specific and invariant features based on their respective weights. Overall, we analyze the characteristics of various features and propose modality modulation strategies that mitigate modality imbalance. Extensive experiments on two multimodal sentiment analysis datasets, demonstrate the superior performance of our method.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".