Social information is widely used in recommender systems to alleviate data sparsity. Since users play a central role in both user-user social graphs and user-item interaction graphs, many previous social recommender systems model the information diffusion process in both graphs to obtain high-order information. We argue that this approach does not explicitly encode high-order connectivity, resulting in potential collaborative signals between user and item not being captured. Moreover, direct modeling of explicit interactions may introduce noises into the model and we expect users to pay more attention to reliable links. In this work, we propose a new recommendation framework named Metapath Enhanced Lightweight Graph Neural Network (ME-LGNN), which fuses social graphs and interaction graphs into a unified heterogeneous graph to encode high-order collaborative signals explicitly. We consider using a lightweight GCN to model collaborative signals. To enable users to capture reliable information more efficiently, we design meta-paths to further enhance the embedding learning by calculating meta-path dependency probabilities. Empirically, we conduct extensive experiments on three public datasets to demonstrate the effectiveness of our model.

论文针对答案选择任务进行研究,利用深层神经网络结合外部知识库信息的方法,提出基于知识库关系信息的双向长短时记忆网络(Knowledge Based Relation-Bidirectional Long Short Term Memory,KBR-BiLSTM),引入知识库中实体信息及实体关系信息去优化基准模型中的注意力机制;并利用知识库关系信息结合上下文丰富了问答的句子编码信息,提升模型效果.在维基问答(Wiki QA)数据集和TREC QA数据集上进行对比实验,证明了KBR-BiLSTM模型的有效性.

Nowadays consumer loan plays an important role in promoting the economic growth, and credit cards are the most popular consumer loan. One of the most essential parts in credit cards is the credit limit management. Traditionally, credit limits are adjusted based on limited heuristic strategies, which are developed by experienced professionals. In this paper, we present a data-driven approach to manage the credit limit intelligently. Firstly, a conditional independence testing is conducted to acquire the data for building models. Based on these testing data, a response model is then built to measure the heterogeneous treatment effect of increasing credit limits (i.e. treatments) for different customers, who are depicted by several control variables (i.e. features). In order to incorporate the diminishing marginal effect, a carefully selected log transformation is introduced to the treatment variable. Moreover, the model's capability can be further enhanced by applying a non-linear transformation on features via GBDT encoding. Finally, a well-designed metric is proposed to properly measure the performances of compared methods. The experimental results demonstrate the effectiveness of the proposed approach.

In this paper, we study the problem of multi-choice reading comprehension, which requires a machine to select the correct answer from a set of candidates based on the given passage and question. Most existing approaches focus on designing sophisticated attention to model the interactions of the sequence triplets (passage, question and candidate options), which aims to extract the answer clues from the passage. After this matching stage, a simple pooling operation is usually applied to aggregate the matching results to make final decisions. However, a bottom-up max or average pooling may loss essential information of the evidence clues and ignore the inter relationships of the sentences, especially dealing with complex questions when there are multiple evidence clues. To this end, we propose an option attentive capsule network with dynamic routing to overcome this issue. Instead of pooling, we introduce a capsule aggregating layer to dynamically fuse the information from multiple evidence clues and iteratively refine the matching representation. Furthermore, we design an option attention-based routing policy to focus more on each candidate option when clustering the features of low-level capsules. Experimental results demonstrate that our proposed model achieves state-of-the-art performance on RACE dataset.

We propose a multi-granularity co-reasoning model for multi-choice reading comprehension task, which aims to select the correct option based on the interaction between passage, question and candidate options. Firstly, we introduce a multiple granularity text matching module to interact passage with question and each option. We take advantage of information extracted from diverse semantic spaces to conduct more extensive matching between text sequences. With this help, we could better match the passage against the question and each option to gather relevant information. Furthermore, we employ a multi-sentence co-reasoning module for sentence inference across multiple sentences. Specifically, we utilize 1D Convolutional Neural Network (1D-CNN) with different kernel sizes and self-attentive Recurrent Neural Network (RNN) to model the relationships of relevant sentences. This module could better synthesize and aggregate sentence-level evidence to make decisions. Experimental results demonstrate that our proposed model achieves state-of-the-art performance for single models on the RACE dataset.

We propose an end-to-end multimodal deep memory network to integrate image and text information for visual-textual sentiment analysis. We focus on the vital regions of an image based on the corresponding text representation. The image regions are treated as memory cells stored in the memory and attention mechanism is leveraged to retrieve relevant visual regions. For feature fusion, we utilize Convolutional Neural Network (CNN) to combine visual and textual information, which is more efficient in learning the joint representation. Experimental results on IMDB dataset demonstrate the effectiveness of our approach.

Detecting communities in networks is to find subgroups of nodes with similar characteristics, which is commonly defined as finding groups of nodes with dense connection in undirected networks. However, communities in directed networks can represent connectivity patterns because of asymmetric relations' which is difficult to capture using traditional algorithms. In this paper, a Gamma-Poisson block model is proposed for community detection in directed networks, which can model not only assortative communities but also communities with various connectivity patterns due to a block matrix. The model can also be extended to undirected networks if we set the block matrix symmetric, and for assortative community detection task if we set the block matrix diagonal. We develop an efficient Gibbs sampling algorithm for the inference work, which can scale to large sparse networks since links other than node pairs are considered during each iteration. We compare our model with several previous ones on a variety of real-world networks and the results demonstrate the advantages in our model.