Aspect-level sentiment classification aims to predict the sentiment polarities towards the target aspects given in sentences. To address the issues of insufficient semantic information extraction and high computational complexity of attention mechanisms in existing aspect-level sentiment classification models based on deep learning, a contextual graph attention network (CGAT) is proposed. The proposed model adopts two graph attention networks to aggregate syntactic structure information into target aspects and employs a contextual attention network to extract semantic information in sentence-aspect sequences, aiming to generate aspect-sensitive text features. In addition, a syntactic attention mechanism based on syntactic relative distance is proposed, and the Gaussian function is cleverly introduced as a syntactic weight function, which can reduce computational complexities and effectively highlight the words related to aspects in syntax. Experiments on three public sentiment datasets show that the proposed model can make better use of semantic information and syntactic structure information to improve the accuracy of sentiment classification.

Accurate identification of weed species is a fundamental task for promoting efficient farmland management. Existing recognition approaches are typically based on either conventional Convolutional Neural Networks (CNNs) or the more recent Vision Transformers (ViTs). CNNs demonstrate strong capability in capturing local spatial patterns, yet they are often limited in modeling long-range dependencies. In contrast, ViTs can effectively capture global contextual information through self-attention, but they may neglect fine-grained local features. These inherent shortcomings restrict the recognition performance of current models. To overcome these limitations, we propose a lightweight hybrid architecture, termed RepEfficientViT,which integrates convolutional operations with Transformer-based self-attention. This design enables the simultaneous aggregation of both local details and global dependencies. Furthermore, we employ a structural re-parameterization strategy to enhance the representational capacity of convolutional layers without introducing additional parameters or computational overhead. Experimental evaluations reveal that RepEfficientViT consistently surpasses state-of-the-art CNN and Transformer baselines. Specifically, the model achieves an accuracy of 94.77%, a precision of 94.75%, a recall of 94.93%, and an F1-score of 94.84%. In terms of efficiency, RepEfficientViT requires only 223.54 M FLOPs and 1.34 M parameters, while attaining an inference latency of merely 25.13 ms on CPU devices. These results demonstrate that the proposed model is well-suited for deployment in edge-computing scenarios subject to stringent computational and storage constraints.

In agricultural planting, early plant classification is an indicator of crop health and growth. In order to accurately classify early plants, this paper proposes a classification method combining a dual attention mechanism and multi-scale module. Firstly, the ECA module (Efficient channel attention) is added to enhance the attention of the network to plants and suppress irrelevant background noise; secondly, the MSFN (Multi-scale Feedforward Network) module is embedded to improve the ability to capture complex data features. Finally, CA (Channel attention) is added to further emphasize the extracted features, thus enhancing the discrimination ability and improving the accuracy of the model. The experimental results show an accuracy of 93.20%, precision of 94.53%, recall of 93.27%, and an F1 score of 93.39%. This study can realize the classification of early plants, and effectively distinguish crops from weeds, which is helpful to identify and realize accurate weeding, thus promoting the intelligent and modern process of agricultural production.

With the rapid expansion of the global fish maw industry, the increasing prevalence of counterfeit products has made authenticity detection a critical challenge. Traditional detection methods rely on organoleptic assessment, chemical analysis, or molecular techniques, which limits their practical application. This paper treats fish maw authenticity detection as an object detection problem and proposes RDT-YOLO, a lightweight detection algorithm based on YOLO11n. Specifically, to address the challenges of insufficient fine texture feature extraction and computational redundancy in fish maw detection, we design hierarchical reparameterized feature extraction modules that utilize reparameterization technology to enhance texture feature extraction capability at different scales. To mitigate information loss during multi-scale feature fusion, we develop a Dynamic Adaptive Multi-Scale Pyramid Processing (DAMSPP) module that incorporates dynamic convolution mechanisms for adaptive feature aggregation. Additionally, we propose an Adaptive Task-Aligned Detection Head (ATADH) that combines task interaction and shared convolution to reduce model parameters while improving detection accuracy. Furthermore, a Wise-ShapeIoU loss function is introduced by incorporating a focusing coefficient into Shape-IoU, enhancing model detection performance through improved bounding box shape optimization. Experimental validation demonstrates that RDT-YOLO achieves 91.9% precision, 89.6% recall, and 94% mAP@0.5 while reducing parameters, model size, and computational complexity by 75.6%, 73.8%, and 63.8%, respectively, compared to YOLO11s. When evaluated against YOLOv10s and YOLOv12s, RDT-YOLO shows mAP@0.5 improvements of 0.8% and 0.5%, respectively. This work provides an automated solution for fish maw authenticity detection with potential for broader food safety applications.