Search for: All records

Multimodal Sentiment Analysis (MSA) leverages heterogeneous modalities, such as language, vision, and audio, to enhance the understanding of human sentiment. While existing models often focus on extracting shared information across modalities or directly fusing heterogeneous modalities, such approaches can introduce redundancy and conflicts due to equal treatment of all modalities and the mutual transfer of information between modality pairs. To address these issues, we propose a Disentangled-Language-Focused (DLF) multimodal representation learning framework, which incorporates a feature disentanglement module to separate modality-shared and modality-specific information. To further reduce redundancy and enhance language-targeted features, four geometric measures are introduced to refine the disentanglement process. A Language-Focused Attractor (LFA) is further developed to strengthen language representation by leveraging complementary modality-specific information through a language-guided cross-attention mechanism. The framework also employs hierarchical predictions to improve overall accuracy. Extensive experiments on two popular MSA datasets, CMU-MOSI and CMU-MOSEI, demonstrate the significant performance gains achieved by the proposed DLF framework. Comprehensive ablation studies further validate the effectiveness of the feature disentanglement module, language-focused attractor, and hierarchical predictions. 

Abstract TheH‐κmethod (Zhu & Kanamori, 2000,https://doi.org/10.1029/1999JB900322) has been widely used to estimate the crustal thickness (H) and the ratio ofPtoSvelocities (V_P/V_Sratio,κ) with receiver functions. However, in regions where the crustal structure is complicated, the method may produce biased results, arising particularly from dipping Moho and/or crustal anisotropy.H‐κstacking in case of azimuthal or radial anisotropy with flat Moho has been proposed but not for cases with plunging anisotropy and dipping Moho. Here we propose a generalizedH‐κmethod calledH‐κ‐c, which corrects for these effects first before stacking. We consider rather general cases, including plunging anisotropy and dipping interfaces of multiple layers, and use harmonic functions to correct for arrival time variations ofPsand its crustal multiples with back azimuth (θ). Systematic synthetic tests show that the arrival time variations can be well fitted by cosθand cos2θfunctions even for very complex crustal structures. Correcting for the back azimuthal variations significantly enhancesH‐κstacking. We verify the feasibility of theH‐κ‐c method by applying it to 40 permanent stations in various geological setting across the Mainland China. The results show clear improvement after the harmonic corrections, with clearer multiples and stronger stacking energy, as well as more reliableH‐κvalues. Large differences inH(up to 5.0 km) andκ(up to 0.09) between the new and traditional methods occur mostly in mountainous regions, where the crustal structure tends to be more complex. We caution in particular about systematic bias when the traditional method is used in the presence of dipping interfaces. The modified method is simple and applicable anywhere in the world. 

Abstract Fluid‐like sliding graphenes but with solid‐like out‐of‐plane compressive rigidity offer unique opportunities for achieving unusual physical and chemical properties for next‐generation interfacial technologies. Of particular interest in the present study are graphenes with specific chemical functionalization that can predictably promote adhesion and wetting to substrate and ultralow frictional sliding structures. Lubricity between stainless steel (SS) and diamond‐like carbon (DLC) is experimentally demonstrated with densely functionalized graphenes displaying dynamic intersheet bonds that mechanically transform into stable tribolayers. The macroscopic lubricity evolves through the formation of a thin film of an interconnected graphene matrix that provides a coefficient of friction (COF) of 0.01. Mechanical sliding generates complex folded graphene structures wherein equilibrated covalent chemical linkages impart rigidity and stability to the films examined in macroscopic friction tests. This new approach to frictional reduction has broad implications for manufacturing, transportation, and aerospace.