ABSTRACT. The purpose of this study is to explore the covariation between emergent social events and different types of public emotion. We first extract the indexes of events and emotions from data of Baidu Index. Then the correlation between emergent social events and public emotions and the correlation between different types of public emotions are analyzed by the Granger causality test. Results show that the public's attention to emergent social events involved in this study has gone through a process from explosion, the decline to fading. Moreover, the trajectory of fear in part of the period is consistent with the life cycle model. We also find that fear and anger are the main emotional reactions at the beginning of most emergent social events, and after the attention to the event subsides, sadness and depression increase, and depression has peaked to varying degrees at different stages of the event. Finally, Granger analyses show that most of the event indexes can predict fear in short-term, while they can predict sadness and depression in the middle and long term. Also, fear and anger can more or less predict subsequent sadness and depression. The results could verify the individual response to stress at the group level.

ABSTRACT. Fake news detection has recently become an important topic of research. This is due to the impact of fake news on the internet specially on social media. The last United States (US)’ presidential election has revealed that fake news can have a political and social impact. Numerous of the models proposed in the previous studies are based on supervised learning. Therefore, these models are unable to deal with the huge amount of unlabeled data about fake news. Few researches focused on early detection. In this study, we built a semi-supervised learning model to detect fake news on social media at an early stage. By using a semi-supervised learning, we make our model able to deal with the huge amount of unlabeled data on social media. We first built a model to extract users’ opinion expressed in comments using LSTM units and GloVe as Embedding layer. Then we used CredRank Algorithm to evaluate users’ credibility using their characteristics. Then, we built a small network of users involved in the spread of a given news. The outputs of these three steps serve as inputs of our news classifier SSLNews. SSLNews is composed of three networks: a shared CNN, an unsupervised CNN and a supervised CNN. We used real world datasets to evaluate our model, Politifact and Gossipcop. When using 25% of labeled data, SSLNews reaches an accuracy of 72.25% on Politifact and 70.35% on Gossipcop. When using data produced in the first 10 minutes of the beginning of the spread of the news, SSLNews reaches an accuracy of 71.10% on Politifact and 68.07% on Gossipcop.

ABSTRACT. Scientific collaboration plays an important role in modern academic research. Collaborations between scholars will bring about high-quality papers and improve the academic influence of scholars. However, it is more and more difficult to find a suitable collaborator due to the rapid growth of academic data. There are already some recommendation systems based on calculating the similarity between scholars. But most of them do not consider the dynamic nature of the scientific collaboration network. To this end, we propose a collaborator recommendation algorithm based on dynamic attribute network representation learning (DANRL). It takes advantage of the network topology, scholar attributes and the dynamic nature of the network to represent scholars as low-dimensional vectors. By calculating the cosine similarity between scholar vectors, we can recommend the most similar collaborators to target scholars. Moreover, at each time step of the dynamic network, our method only needs to train embedding vectors for some selected nodes instead of performing random walks and training embedding vectors for all nodes, which can significantly improve the recommendation efficiency. Experiments on two real-world datasets show that DANRL outperforms several baseline methods.

ABSTRACT. In recent years, various types of network attacks emerge in endlessly, the protection of network security has been paid more and more attention by our society. Network Intrusion Detection System (NIDS)is used to protect computer systems from malicious attacks and intrusions, thus has also become a hot research field. Due to the great success of deep learning in industry and academia, there is an increasing interest in the application of deep learning methods for feature representations and classification. In this paper,we propose a intrusion detection model based on time-related deep learning approach with attention mechanism. Firstly, we build a stacked sparse autoencoder(SSAE) to extract high-level feature representations of intrusion information. Then we design a two-layer bidirectional gated recurrent unit(BiGRU) network with attention mechanism to classify traffic data. We perform experiments on a benchmark dataset UNSW-NB15, the results in binary classification indicate that using high-dimensional sparse features extracted by SSAE can significantly accelerate the classification progress. Our model can detect network intrusions effectively and outperform other related methods with reduction of false alarm rate, high accuracy rate, reduced training and testing time.

ABSTRACT. Knowledge base question answering(KBQA) is the key technology of natural language processing. How to understand the semantic information of the natural language problem and capture the semantic relationship between the problem and the structured triples are the problems that KBQA needs to solve. The boundary of subject entities in Chinese questions is not as clear as English, which increases the difficulty of entity recognition. Besides, the variable Chinese grammar makes predicate mapping more difficult for semantic analysis. Existing KBQA is usually implemented using a pipeline model, which has two disadvantages: (1) Errors caused by entity recognition will be propagated to predicate mapping. (2) Neither entity recognition nor predicate mapping can benefit from the information available to each other. So we propose a BERT-based KBQA to joint entity recognition and predicate mapping tasks that use their dependencies to improve model performance. BERT can solve the semantic ambiguity of the Chinese Q&A databases and improve the accuracy of Chinese Knowledge Base Question Answering(CKBQA). The model achieved an F1 score of 92.04% on the NLPCC 2016 KBQA dataset.

ABSTRACT. Public procurement is a business process that is prone to corruption and administrative inefficiency, affecting quality of service delivery to the public. Using Bicol University’s three-year procurement data as a sample, this paper explores the use of process mining on publicly-available procurement data to discover underlying structure of procurement processes of government entities in the Philippines, check for conformance with the prescribed process in the procurement law, and identify potentially problematic nodes. In this report, event logs were generated from official public procurement data and mined with heuristics-based process mining algorithm, using free, open-sourced tools. The discovered processes revealed a concept drift in publication of award, a point for inspection and improvement for the agencies involved.

ABSTRACT. The study considered data from EEG rhythms in the eyes-closed at rest and the eyes-open condition during dynamic movements in real-time soccer training. The EEG was recorded in the orbitofrontal cortex using NeuroSky single-channel wireless mobile system with pair dry non-contact sensors. The participants included professional male soccer players. Results from this study showed reducing the power spectrum of EEG rhythms during soccer training compared to rest condition and demonstrated statistically significant differences (p<0.03) between the rest and during dynamic movement conditions obtained as the summary value of bands in EEG power spectral estimates (1-50 Hz). The decrease the power spectrum in frontal areas associate with “neural efficiency” among team sports athletes and relationship to cognitive function as well. In addition, the findings are interpreted to suggest that delta rhythm are a plausible neurobiological index of physical fatigue during sport training among soccer players. These findings encourage application of wireless portable EEG systems for the studies of brain functions among sportspersons.

ABSTRACT. eSports has tremendously evolved within the last two decades. Myriads of competitions and tournaments are available for amateur and professional players. A pro-player is typically exposed to the stress situations during the competitions with huge number of visitors, critical in-game situations as well as long and tedious trainings. Pro-eSports teams used to engage psychologists for helping the players to overcome the psychological situations. However, there is a distinct lack of research on stressors and coping strategies in eSports which reduces the effectiveness of psychological assistance. The purpose of this study is to characterize the pro-eSports team players regarding their personality traits and coping strategies. We collected the data from pro-players of different teams, as well as from amateur players, and found significant discrepancies in coping strategies between the professional and amateur gamers. The most distinctive features of pro-players appeared to be a frequent use of the strategy of seeking social support and reducing the self-control in a difficult situation.

ABSTRACT. Connected Autonomous Vehicles (CAVs) are drive-less vehicles that equipped with sophisticated technologies that provide safe and efficient transportation for the users. The automated and connected nature of a CAV make them vulnerable to cyber attack. Many researchers proposed various solutions for establishing security in the communication. Majority focuses on security services such as confidentiality, integrity, availability and authentication. We analyzed the most recent concepts that have been discussed in the literature to protect the CAVs from the attackers. We then introduce our new security framework that can address the gaps in the current security knowledge by extending our previous ideas and solutions.