Human-computer interaction under the cloud computing platform is very important, but the semantic gap will limit the performance of interaction. It is necessary to understand the semantic information in various scenarios. Relation classification (RC) is an import method to implement the description of semantic formalization. It aims at classifying a relation between two specified entities in a sentence. Existing RC models typically rely on supervised learning and distant supervision. Supervised learning requires large-scale supervised training datasets, which are not readily available. Distant supervision introduces noise, and many long-tail relations still suffer from data sparsity. Few-shot learning, which is widely used in image classification, is an effective method for overcoming data sparsity. In this paper, we apply few-shot learning to a relation classification task. However, not all instances contribute equally to the relation prototype in a text-based few-shot learning scenario, which can cause the prototype deviation problem. To address this problem, we propose context attention-based prototypical networks. We design context attention to highlight the crucial instances in the support set to generate a satisfactory prototype. Besides, we also explore the application of a recently popular pre-trained language model to few-shot relation classification tasks. The experimental results demonstrate that our model outperforms the state-of-the-art models and converges faster.

Internet of Thing (IoT) systems have been treated as a novel platform for graph data acquisition. Contents like dynamic network topology, organization and control flows, and interactions among monitored objects all contribute to the huge volumes of graph data generated in IoT. These data are believed to brought significant benefits to both the operation and functionalities of IoT systems, especially when combined with cutting-edge Artificial Intelligence techniques. However, these graph data are usually locally collected by data contributors with sensing devices, which could be both partially overlapped as they record same environment, and sensitive as they can indicate private physical status of contributors. Considering all challenges, current solutions for graph data collection in IoT are incapable. Therefore, this article proposes a novel framework for privacy-preserving distributed graph data collection for IoT. The framework allows the graphs kept by data contributors to be partially overlapped, and can help the data broker to efficiently derive the universal view by combining these graphs. The differential privacy is applied for privacy preservation during data collection. The proposed problem aims at minimizing the total bandwidth consumption for graph collection, which is proved to be NP-complete. Then three algorithms are proposed for different circumstances, based on the diverse knowledge and purposes held by the data broker. Finally, both theoretical and numerical analysis have demonstrated the advancement of these methods.

Graph neural networks have been proven to be very effective for representation learning of knowledge graphs. Recent methods such as SACN and CompGCN, have achieved the most advanced results in knowledge graph completion. However, previous efforts mostly rely on localized first-order approximations of spectral graph convolutions or first-order neighborhoods, ignoring the abundant local structures like cycles and stars. Therefore, the diverse semantic information beneath these structures is not well-captured, leaving opportunities for better knowledge representation which will finally help KGC. In this work, we propose LSA-GAT, a graph attention network with a novel neighborhood aggregation strategy for knowledge graph completion. The model can take special local structures into account, and derive a sophisticated representation covering both the semantic and structural information. Moreover, the LSA-GAT model is combined with a CNN-based decoder to form an encoder-decoder framework with a carefully designed training process. The experimental results show significant improvement of the proposed LSA-GAT compared to current state-of-the-art methods on FB15k-237 and WN18RR datasets.

Medical image segmentation plays an important role in clinical diagnosis, quantitative analysis, and treatment process. Since 2015, U-Net-based approaches have been widely used for medical image segmentation. The purpose of the U-Net expansive path is to map low-resolution encoder feature maps to full input resolution feature maps. However, the consecutive deconvolution and convolutional operations in the expansive path lead to the loss of some high-level information. More high-level information can make the segmentation more accurate. In this paper, we propose MU-Net, a novel, multi-path upsampling convolution network to retain more high-level information. The MU-Net mainly consists of three parts: contracting path, skip connection, and multi-expansive paths. The proposed MU-Net architecture is evaluated based on three different medical imaging datasets. Our experiments show that MU-Net improves the segmentation performance of U-Net-based methods on different datasets. At the same time, the computational efficiency is significantly improved by reducing the number of parameters by more than half.