With the rapid growth of data, knowledge base question answering (KBQA) is becoming more and more important. However, most existing methods of KBQA take every word in the question into account, leading to serious semantic confusion of the question and low efficiency of the question answering (QA) System. Therefore, we proposed an intent identification method to wipe off irrelevant words to decrease the semantic influence. By locating, expanding and disambiguating the subject and its attributes of questions, we not only obviously decrease the time cost of KBQA but also greatly reduce the amount of data processing and search space. Furthermore, by incorporating Convolutional Neural Networks (CNNs) to model questions and answer candidates, the top ranking candidates can be easily identified as answers. Experiments on an well-known track of NLPCC 2016 dataset show that the average F1 score is 75.26%, which is much higher than previous methods.
Vehicular ad hoc network (VANET) has been brought out to revolutionize driving experiences and traffic flow control systems. In a VANET, roadside units (RSUs) are usually used to collect messages sent by nearby vehicles, which may include status information of vehicles and emergency events. Currently, batch verification is considered as an effective approach to verify these messages in VANETs. However, in existing batch verification scheme, the batch verification fails even one of the signed messages is invalid. In this paper, we introduce an efficient identity-based fault-tolerant batch verification scheme for cloud assisted VANETs.
Signed graphs are prevalent data structures containing both positive and negative links. Recently, the fundamental network analysis task on signed graphs, namely link sign prediction, has received careful attention. Existing methods learn two target node representations independently, and the sign between these two nodes is predicted based on similarity. However, such a paradigm is node-centric that cannot distinguish node pairs with distinct contexts, thus lowering the prediction performance. Learning pair-centric representation is therefore a rewarding way to be aware of differences between pairs. There is no study yet on how to build such an appropriate representation that can effectively infer the sign between the target node pair. In this paper, we provide a new perspective to conduct link sign prediction within the paradigm of subgraph classification and propose a novel Subgraph-based link Sign Prediction (SSP) model. Technically, SSP uses importance-based sampling to extract an informative subgraph around each target node pair. For each subgraph, an innovative node labeling scheme is designed to encode its structural and signed information for representation learning. To further utilize the subgraph representation for imbalanced sign classification, SSP employs self-pruning contrastive learning to gain balanced representations. Extensive experiments on real-world datasets demonstrate that SSP consistently and significantly outperforms all the state-of-the-art baselines.