In this study, we analyzed Chinese traditional fermented food to isolate and identify new lactic acid bacteria (LAB) strains with novel functional properties and to evaluate their cellular antioxidant and bile salt hydrolase (BSH) activities in vitro. A sequential screening strategy was developed to efficiently isolate and obtain 261 LAB strains tolerant of bile salt, acid, and H2O2 from nine Chinese traditional fermented foods. Among these strains, 70 were identified as having 2,2-diphenyl-1-picrylhydrazyl radical scavenging and/or BSH activity. These strains belonged to eight species: Enterococcus faecium (33% of the strains), Lactobacillus plantarum (26%), Leuconostoc mesenteroides (14%), Pediococcus pentosaceus (6%), Enterococcus durans (9%), Lactobacillus brevis (9%), Pediococcus ethanolidurans (3%), and Lactobacillus casei (1%). The pulsed-field gel electrophoresis genome fingerprinting profiles of these strains revealed 38 distinct pulsotypes, indicating a high level of genomic diversity a...
Visual Dialog is a challenging vision-language task since the visual dialog agent needs to answer a series of questions after reasoning over both the image content and dialog history. Though existing methods try to deal with the cross-modal understanding in visual dialog, they are still not enough in ranking candidate answers based on their understanding of visual and textual contexts. In this paper, we analyze the cross-modal understanding in visual dialog based on the vision-language pre-training model VD-BERT and propose a novel approach to improve the cross-modal understanding for visual dialog, named ICMU. ICMU enhances cross-modal understanding by distinguishing different pulled inputs (i.e. pulled images, questions or answers) based on four-way contrastive learning. In addition, ICMU exploits the single-turn visual question answering to enhance the visual dialog model's cross-modal understanding to handle a multi-turn visually-grounded conversation. Experiments show that the proposed approach improves the visual dialog model's cross-modal understanding and brings satisfactory gain to the VisDial dataset.
Real-time, high-precision significant wave height (SWH) fields are essential for ocean forecasting, disaster mitigation, and marine applications. However, extremely sparse observational coverage in vast ocean regions limits accurate reconstruction, as traditional spatial interpolation often produces overly smoothed fields that fail to capture fine-scale structures and sharp gradients. This study proposes the Time-to-Space Wave (T2S-W) Framework, a hybrid approach based on “filling space with time.” It advects historical time-series observations forward to the target time using group velocities to generate virtual observations that augment spatial coverage. A lightweight LightGBM model approximates the net effect of nonlinear source and sink processes, applying corrections for dissipation and energy growth along ray paths without real-time wind fields. A convolutional neural network (CNN)–Transformer architecture then performs high-precision continuous reconstruction. Results show strong accuracy and robustness under ~99.8% missing rates, achieving 0.20 m RMSE on test sets while preserving fine-scale structures across monsoon and typhoon conditions. Diagnostics reveal the framework estimates wind-sea/swell decoupling, shallow-water topographic dissipation, and propagation-direction-dependent spatial response patterns. This provides a novel pathway for high-fidelity reconstruction under sparse data and indicates that the potential extends to other advection-diffusion ocean fields like temperature, salinity, and tracers.