Webpage segmentation is the basic building block for a wide range of webpage analysis methods. The rapid development of Web technologies results in more dynamic and complex webpages, which brings new challenges to this area. To improve the performance of webpage segmentation, we propose a two-stage segmentation method that can combine visual, logic and semantic features of the contents on a webpage. This two-stage method can effectively conduct webpage segmentation on complicated and dynamic webpages. The experiment results show that the proposed method significantly outperforms the existing state-of-the-art in terms of higher precision, recall and accuracy.
Online social network services (OSNSs) are changing the fabric of our society, impacting almost every aspect of it. Over the last decades, multiple competing OSNSs have emerged. As a result, users are trapped in the walled gardens of their OSNS, encountering restrictions about the people they can interact with. Our work aims at enabling users to meet and interact beyond the boundary of their OSNSs. We introduce USNB -Universal Social Network Bus which revisits the "service bus" paradigm to address the requirements of social interoperability.
We propose PTP-MF (Pairwise Trust Prediction through Matrix Factorization), an algorithm to predict the intensity of trust/distrust relations in Online Social Networks. The PTP-MF algorithm maps each user i onto two low-dimensional vectors, namely, the trustor profile (describing her/his inclination to trust others) and the trustee profile (modelling how others perceive i as trustworthy) and computes the trust j places in j as the dot product of trustor profile of i and the trustee profile of j. Experiments indicate that the PTP-MF algorithm is more accurate than the state-of-the-art approaches (up to 9.65%) and scales well on real life graphs
In this paper, we propose a Threat Management System (TMS) for Data-driven Internet-of-Things-based Collaborative Systems (DIoTCS). The novelty of the system is that it con nes the damage into partitions generated for the global dataset shared by the DIoTCS applications. We formulate the partitioning problem as a cost-driven optimization problem, which is proven to be NP-hard. Accordingly, two heuristics are proposed to solve this problem. For TMS, we propose response and recovery subsystems. We evaluate TMS experimentally and demonstrate that intelligent partitioning of global dataset improves the overall availability of the DIoTCS.
In this paper, we propose an automatic labeling approach to distinguish crisis-relevant tweets while differentiating source types (e.g., government or personal accounts) simultaneously. We first analyze and identify tweet-specific linguistic, sentimental and emotional features based on statistical topic modeling. Then,we design a novel correlative convolutional neural network which uses a shared hidden layer to learn effective representations of the multi-faceted features. The model can discover salient information robust to the variations and noises in tweets and sources. To obtain a bird-view of crisis event, we further develop an approach to automatically summarize key information of identified tweets.
Previous work in social network de-anonymization mainly focuses on designing accurate and efficient de-anonymization methods. We attempt to investigate the intrinsic relation between the attacker's knowledge and the expected de-anonymization gain. A common intuition is that more knowledge results in more successful de-anonymization. However, our analysis shows their relation is much more sophisticated than that. Though based on a few assumptions, our findings leave intriguing implications for the attacker to make better use of the background knowledge when performing de-anonymization, and for the data owners to better measure the privacy risk when releasing their data to third parties.
In Attribute-Based Access Control (ABAC), access to resources is given based on attributes of subjects, objects, and environment. There is an imminent need for the development of efficient algorithms that enable migration to ABAC. However, existing policy mining approaches do not consider possible adaptation to the policy of a similar organization. In this article, we address the problem of automatically determining an optimal assignment of attribute values to subjects for enabling the desired accesses to be granted while minimizing the number of ABAC rules used by each subject. We show the problem to be NP-Complete and propose a heuristic solution.
This study empirically investigates the impact of past experience on human trust in and reliance on agent teammates. We developed a repeated team coordination game, in which two players repeatedly cooperate to complete team tasks. The results show that positive (negative) past experience increases (decreases) human trust in and reliance agent teammates; lack of past experience lead to higher trust levels compared to positive past experience. These findings provide clear and significant evidence and enhance our understanding of the changes in human trust in peer-level agent teammates with respect to past experience
We focus on using unstructured textual Knowledge Bases (KBs) to answer questions from community based Question-and-Answer (Q&A) websites. We propose a novel framework that integrates multi-level tag recommendation with external KBs to retrieve the most relevant KB articles to answer user posted questions. A post-tag co-clustering model, augmented by a two-step tag recommender, is used to predict tags at different levels for a given user posted question. A KB article retrieval component leverages the recommended multi-level tags to select the appropriate KBs and search/rank the matching articles thereof. We conduct extensive experiments to demonstrate the effectiveness of the proposed approach.
Search engines cannot answer time and location-specific queries; this often requires humans on-site. While community question answering (CQA) platforms are popular, few exceptions consider users physical locations. Here, we present CloseUp, our prototype for the integration of community-driven live search into a Google-like search experience. We bridge the gap between Web search and CQA, namely the formulation of search requests and the expected response times. CloseUp features a deep learning pipeline to translate relevant queries into questions. CloseUp provides a mobile application for submitting and replying to questions. Using a field study, we evaluated the feasibility of our approach.