Machine reading comprehension (MRC), which requires a machine to answer questions based on a given context, has attracted increasing attention with the incorporation of various deep-learning techniques over the past few years. Although research on MRC based on deep learning is flourishing, there remains a lack of a comprehensive survey summarizing existing approaches and recent trends, which motivated the work presented in this article. Specifically, we give a thorough review of this research field, covering different aspects including (1) typical MRC tasks: their definitions, differences, and representative datasets; (2) the general architecture of neural MRC: the main modules and prevalent approaches to each; and (3) new trends: some emerging areas in neural MRC as well as the corresponding challenges. Finally, considering what has been achieved so far, the survey also envisages what the future may hold by discussing the open issues left to be addressed.
This paper focuses mainly on the problem of Chinese medical question answer matching, which is arguably more challenging than open-domain question answer matching in English due to the combination of its domain-restricted nature and the language-specific features of Chinese. We present an end-to-end character-level multi-scale convolutional neural framework in which character embeddings instead of word embeddings are used to avoid Chinese word segmentation in text preprocessing, and multi-scale convolutional neural networks (CNNs) are then introduced to extract contextual information from either question or answer sentences over different scales. The proposed framework can be trained with minimal human supervision and does not require any handcrafted features, rule-based patterns, or external resources. To validate our framework, we create a new text corpus, named cMedQA, by harvesting questions and answers from an online Chinese health and wellness community. The experimental results on the cMedQA dataset show that our framework significantly outperforms several strong baselines, and achieves an improvement of top-1 accuracy by up to 19%.
Due to the existence of information overload in social networks, it becomes increasingly difficult for users to find useful information according to their interests. This paper takes Twitter-like social networks into account and proposes models to characterize the process of information diffusion under information overload. Users are classified into different types according to their in-degrees and out-degrees, and user behaviors are generalized into two categories: generating and forwarding. View scope is introduced to model the user information-processing capability under information overload, and the average number of times a message appears in view scopes after it is generated by a given type user is adopted to characterize the information diffusion efficiency, which is calculated theoretically. To verify the accuracy of theoretical analysis results, we conduct simulations and provide the simulation results, which are consistent with the theoretical analysis results perfectly. These results are of importance to understand the diffusion dynamics in social networks, and this analysis framework can be extended to consider more realistic situations.
Frequent subgraph mining (FSM) plays an important role in graph mining, attracting a great deal of attention in many areas, such as bioinformatics, web data mining and social networks. In this paper, we propose SSiGraM (Spark based Single Graph Mining), a Spark based parallel frequent subgraph mining algorithm in a single large graph. Aiming to approach the two computational challenges of FSM, we conduct the subgraph extension and support evaluation parallel across all the distributed cluster worker nodes. In addition, we also employ a heuristic search strategy and three novel optimizations: load balancing, pre-search pruning and top-down pruning in the support evaluation process, which significantly improve the performance. Extensive experiments with four different real-world datasets demonstrate that the proposed algorithm outperforms the existing GraMi (Graph Mining) algorithm by an order of magnitude for all datasets and can work with a lower support threshold.
Unmanned aerial vehicles (UAVs) are widely applied in smart city applications such as urban sensing and delivery, due to the UAVs’ agility, low cost and not being restricted by ground road conditions. However, the limited battery capacity becomes one of the biggest obstacles to the application of UAVs. To address this issue, this article investigates an emergency response application, in which UAVs generally ride crowdsourced buses to save energy and respond to a stochastic emergency event (such as a traffic accident) when the event occurs. For the bus-based UAV response paradigm, a single UAV response process with the constraint of the bus mobility is first modeled. Subsequently, a data-driven UAV path planning algorithm is designed. Then two emergency response cases by multi-UAV are investigated. One case is irregular emergency response, whose objective is to maximize the temporal-spatial coverage of the urban area. The other case is predictable emergency response, which optimizes the response performance to these emergencies. Thereafter, the bus-stimulating problems for the two cases are formulated and solved. Finally, utilizing a real-world bus trajectory data set generated by a large-scale bus fleet and a traffic event data set, the emergency response performance of the bus-based UAV response paradigm is comprehensively evaluated. The results show that (1) with only 30 UAVs, 90% of Shenzhen city can be covered in the irregular emergency response case; (2) with only 50 UAVs, the average response delay to the emergencies is shorter than 1.5 min, which is 56% shorter than baselines, in the predictable emergencies response case.
The task of Cross-lingual Passage Re-ranking (XPR) aims to rank a list of candidate passages in multiple languages given a query, which is generally challenged by two main issues: (1) the query and passages to be ranked are often in different languages, which requires strong cross-lingual alignment, and (2) the lack of annotated data for model training and evaluation. In this article, we propose a two-stage approach to address these issues. At the first stage, we introduce the task of Cross-lingual Paraphrase Identification (XPI) as an extra pre-training to augment the alignment by leveraging a large unsupervised parallel corpus. This task aims to identify whether two sentences, which may be from different languages, have the same meaning. At the second stage, we introduce and compare three effective strategies for cross-lingual training. To verify the effectiveness of our method, we construct an XPR dataset by assembling and modifying two monolingual datasets. Experimental results show that our augmented pre-training contributes significantly to the XPR task. Besides, we directly transfer the trained model to test on out-domain data which are constructed by modifying three multi-lingual Question Answering (QA) datasets. The results demonstrate the cross-domain robustness of the proposed approach.