Abstract: Disclosed is a dialogue emotion correction method based on a graph neural network, including: extracting acoustic features, text features, and image features from a video file to fuse them into multi-modal features; obtaining an emotion prediction result of each sentence of a dialogue in the video file by using the multi-modal features; fusing the emotion prediction result of each sentence with interaction information between talkers in the video file to obtain interaction information fused emotion features; combining, on the basis of the interaction information fused emotion features, with context-dependence relationship in the dialogue to obtain time-series information fused emotion features; correcting, by using the time-series information fused emotion features, the emotion prediction result of each sentence that is obtained previously as to obtain a more accurate emotion recognition result.

Abstract: A monocrystalline silicon micro-nano dual-scale anti-reflection texture and a preparation method therefor. The preparation method combines nanosecond-laser-assisted waterjet near-damage-free processing and femtosecond laser scanning, and subsurface damage caused by a re-cast layer phenomenon and a hot crack in a monocrystalline silicon laser texturing process can be effectively reduced by combining a nanosecond-laser-assisted waterjet near-damage-free processing technology and an ultra-short pulse femtosecond laser cold processing technology; and meanwhile, a micro-scale frame structure and a nano-scale structure can be flexibly modified respectively by adjusting nanosecond-laser-assisted waterjet technological parameters and femtosecond laser technological parameters, a geometry light trapping effect and an effective dielectric effect can be achieved in a micro-nano dual-scale hybrid structure at the same time, and surface reflection is reduced.

Abstract: The invention relates to lipstick compositions including at least one natural oil, at least one fatty alcohol wax and particles of at least one filler, in which the lipstick compositions contain minimal amount of animal-derived ingredients and/or beeswax, or are free of such ingredients, as well as to methods and kits including such corn positions.

Abstract: Provided herein are improved, cost-effective and environmentally friendly methods of production of recombinant AAV (rAAV) in embryonated avian eggs. Further provided herein is a provides embryonated avian eggs as novel host vehicles for high-yield production of rAAV, including both packaging and propagation. In particular, embryonated chicken eggs provide a novel expression vehicle for AAV of mammalian origin, irrespective of AAV serotype. The disclosed methods may comprise packaging of rAAV in embryonated avian eggs (e.g., chicken eggs) by inoculating an embryonated avian egg with a first nucleic acid vector comprising a transgene and a second nucleic acid vector comprising AAV rep and cap genes, incubating the egg, and isolating rAAV from the egg, wherein the AAV is of non-avian origin. Also provided are methods of purifying and propagating packaged rAAV in embryonated avian eggs or in avian embryonic fibroblasts.

Abstract: Disclosed is a dialogue emotion correction method based on a graph neural network, including: extracting acoustic features, text features, and image features from a video file to fuse them into multi-modal features; obtaining an emotion prediction result of each sentence of a dialogue in the video file by using the multi-modal features; fusing the emotion prediction result of each sentence with interaction information between talkers in the video file to obtain interaction information fused emotion features; combining, on the basis of the interaction information fused emotion features, with context-dependence relationship in the dialogue to obtain time-series information fused emotion features; correcting, by using the time-series information fused emotion features, the emotion prediction result of each sentence that is obtained previously as to obtain a more accurate emotion recognition result.

Abstract: The present invention provides a virulent Pseudomonas fluorescens (P. fluorescens) phage ?Pf1901, and a phage ?Pf1901 preparation and an application thereof, and relates to the technical field of phage. The virulent P. fluorescens phage ?Pf1901 has an accession number of CCTCC M2019447. The virulent phage ?Pf1901 has a titer of (1.4-3)×1010 PFU/mL. The virulent phage ?Pf1901 has an optimal multiplicity of infection (MOI) value of 0.0001. The virulent P. fluorescens phage ?Pf1901 provided by the present invention exhibits very high specificity and lytic ability to P. fluorescens, which can be used to control P. fluorescens, with strong lytic and scavenging effects on a host.

Abstract: The present invention provides a virulent Pseudomonas fluorescens (P. fluorescens) phage ?Pf1901, and a phage ?Pf1901 preparation and an application thereof, and relates to the technical field of phage. The virulent P. fluorescens phage ?Pf1901 has an accession number of CCTCC M2019447. The virulent phage ?Pf1901 has a titer of (1.4?3)×1010 PFU/mL. The virulent phage ?Pf1901 has an optimal multiplicity of infection (MOI) value of 0.0001. The virulent P. fluorescens phage ?Pf1901 provided by the present invention exhibits very high specificity and lytic ability to P. fluorescens, which can be used to control P. fluorescens, with strong lytic and scavenging effects on a host.