SemMSA: Latent Semantic-Aided Robust Multimodal Sentiment Analysis with Incomplete Data
SemMSA reaches state-of-the-art multimodal sentiment analysis on incomplete data using latent LLM semantics.
SemMSA is a multimodal sentiment-analysis framework for incomplete language, visual, and acoustic inputs. Cross-modal Semantic Refinement uses adapters and a frozen LLM to iteratively produce continuous semantic states without decoding text, and Cross-modal Spectral Alignment matches those semantics to every modality via the dominant component of a kernel Gram matrix. An instance-level spectral separation constraint is added to preserve discriminability. Experiments on SIMS, MOSI, and MOSEI are reported as state of the art.
- SemMSA builds sentiment semantics in a frozen LLM without decoding text.
- Spectral alignment matches those semantics to language, vision, and audio.
- An instance-level spectral constraint is used to limit representation collapse.
- The method reports state-of-the-art results on SIMS, MOSI, and MOSEI.
Full article212 words · extracted from arxiv.org · click to collapse
Recent research on Multimodal Sentiment Analysis (MSA) has focused on learning from language, visual, and acoustic modalities with incomplete data to infer human sentiment. Most studies typically compensate for missing information by reconstructing modality features or designing complicated fusion mechanisms. However, these methods still suffer from spurious generation and noisy guidance due to the lack of high-level semantic grounding in partially observed multimodal evidence. To address these issues, we propose SemMSA, a latent semantic-aided framework that constructs rich sentiment-relevant semantics with LLMs, fully integrating with all modalities via anchor-free spectral alignment. It mainly consists of Cross-modal Semantic Refinement (CSR) and Cross-modal Spectral Alignment (CSA). Specifically, CSR first adaptively extracts visual and acoustic representations by corresponding adapters to form a unified multimodal prefix with language in the frozen LLM embedding space. It then iteratively produces continuous discriminative semantic states through a token-efficient latent refinement process without decoding explicit text. Next, CSA simultaneously aligns the refined semantics with all modalities by enhancing the dominant spectral component of their kernel Gram matrix. This captures global nonlinear dependencies among all representations without relying on a predefined anchor modality. In addition, an instance-level spectral separation constraint preserves cross-sample discriminability and mitigates representation collapse. Extensive experiments on SIMS, MOSI, and MOSEI benchmarks demonstrate that SemMSA achieves state-of-the-art performance.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.30238