Search

Dialogue-based multiple-choice machine reading comprehension (MRC) is one of most difficult and novel tasks because it requires more advanced reading comprehension skills, such as speaker’s intention analysis, non-extractive reasoning, commonsense knowledge. Previous models usually only compute attention scores from the fixed representation of entire dialogue, and also do not fully consider the contribution of dialogue, question, options, and their combinations respectively. In this paper, we introduce Utterance-focusing Multiway-matching Network (UMN), a simple but effective human mimicking model for dialogue-based multiple-choice MRC. First, two utterance-focusing mechanisms called ParaUF and AutoUF are proposed to extract the utterances that are most relevant to the question and option: ParaUF gets the bilinear weighted distance between each utterance of dialogue and question and option during training while AutoUF obtains the scores by the relevance, overlap and coverage (ROC) rules before training process. Second, we adopted the multiway-matching mechanism to capture the relationship among the question, option and selected utterances through calculating the attention weights between the quadruplet of four sequences: utterances, question, option and the concatenation of each two. We evaluate the proposed model on dialogue-based multiple-choice MRC tasks, DREAM, and outperformed recently published methods under the same pre-trained model. A series of detailed analysis is also conducted to interpret the differences of two utterance-focusing mechanisms and the effectiveness of the proposed multiway-matching mechanism.

Word embeddings which is real-valued word representation able to capture lexical semantics plays a crucial role in machine reading comprehension tasks because the first step is the embedding of the question and the passage. One of the frequently-used models is Word2Vec, but several fashionable competitors have been proposed in recent years, including GloVe and Fasttext. However the question which word embedding model really performs best and is most suitable across different reading comprehension tasks remains unanswered to this date. In this paper we performed the first extrinsic empirical evaluation of three word embeddings across four types of tasks: Multiple Choice, Cloze, Answer Extraction and Conversation. The experiments showed that GloVe and Fasttext have their own strengths in different type of tasks: The accuracy of Multiple Choice task improves significantly when leveraging GloVe, Fasttext is a bit more suitable for Answer Extraction task, and GloVe performs similarly with Fasttext in Cloze task and Conversation task. Finally, we find that Word2Vec has been demoded compared with GloVe and Fasttext in all tasks.

This paper presents a new secure and targeted mobile coupon delivery scheme based on blockchain. Our goal is to design a decentralized targeted mobile coupon delivery framework, which enables the secure delivery of targeted coupons to eligible mobile users whose behavioral profiles accurately satisfy the targeting profile defined by the vendor. It does not require trusted third-party meanwhile protects the mobile user and vendor’s information security, including user privacy, data integrity and rights protection. We adopt Policy-Data Contract Pair (PDCP) to control the transfer of information between users and vendors and use encryption algorithm to ensure the data security. Once transactions containing signatures are recorded in the blockchain after consensus, they become non-repudiation. Theoretical analysis and simulation experimental results demonstrate that our model has higher security and lower computation than JG’16 scheme.

In computer vision and pattern recognition applications, there are usually a vast number of unlabelled data whereas the labelled data are very limited. Active learning is a kind of method that selects the most representative or informative examples for labelling and training; thus, the best prediction accuracy can be achieved. A novel active learning algorithm is proposed here based on one-versus-one strategy support vector machine (SVM) to solve multi-class image classification. A new uncertainty measure is proposed based on some binary SVM classifiers and some of the most uncertain examples are selected from SVM output. To ensure that the selected examples are diverse from each other, Gaussian kernel is adopted to measure the similarity between any two examples. From the previous selected examples, a batch of diverse and uncertain examples are selected by the dynamic programming method for labelling. The experimental results on two datasets demonstrate the effectiveness of the proposed algorithm.

Intrusion detection is one of the most important problems in network security. Its target is to secure internal networks by identifying unusual access or attacks. Machine learning techniques have been playing a significant role in intrusion detection. Considering the large size of training data and time-consuming labeling task, it is wise to select some informative data to train a classifier. Active learning is a family of approaches selecting samples for labeling to build classifier with maximum prediction accuracy. So it is able to improve the performance of intrusion detection while it is not time-costing and labor-consuming. In this paper, definition and some efficient query strategies of active learning are reviewed and suggested. Some popular algorithms of intrusion detection and the combination of active learning and intrusion detection are also introduced. But existing work of active learning for intrusion detection is very limited. We propose more active learning methods should be developed for intrusion detection.

Active Learning is a machine learning technique that selects the most informative examples for labeling so that the classification performance would be improved to its maximum possibility. In this paper, a novel active learning approach based on Maximum Density and Minimum Redundancy (MDMR) is proposed. The objective of MDMR is to select a set of examples that have large density and small redundancy with others. Firstly, we propose new methods to measure the density and redundancy of examples. Then a model is built to select examples by combining density and redundancy and dynamic programming algorithm is applied to solve the problem. The results of the experiment on terrain classification have demonstrated the effectiveness of the proposed approach.

Semi-supervised learning and active learning are important techniques to solve the shortage of labeled examples. In this paper, a novel active learning algorithm combining semi-supervised Learning with Local and Global Consistency (LLGC) is proposed. It selects the example that can minimize the estimated expected classification risk for labeling. Then, a better classifier can be trained with labeled data and unlabeled data using LLGC. The experiments on two datasets demonstrate the effectiveness of the proposed algorithm.