Attention Based Context Boosted Cyberbullying Detection in Social Media

Cyberbullying detection is a rising research topic due to its paramount impact on social media users, especially youngsters and adolescents. While there has been an enormous amount of progress in utilising efficient machine learning and NLP techniques for tackling this task, recent methods have not fully addressed contextualizing the textual content to the highest possible extent. The textual content of social media posts and comments is normally long, noisy and mixed with lots of irrelevant tokens and characters, and therefore utilizing an attention-based approach that can focus on more relevant parts of the text can be quite pertinent. Moreover, social media information is normally multi-modal in nature and may contain various metadata and contextual information that can contribute to enhancing the Cyberbullying prediction system. In this research, we propose a novel machine learning method that, (i) fine tunes a variant of BERT, a deep attention-based language model, which is capable of detecting patterns in long and noisy bodies of text; (ii)~extracts contextual information from multiple sources including metadata information, images and even external knowledge sources and uses these features to complement the learner model; and (iii) efficiently combines textual and contextual features using boosting and a wide-and-deep architecture. We compare our proposed method with state-of-the-art methods and highlight how our approach significantly outperforming the quality of results compared to those methods in most cases.

Sentiment Mining and Analysis over Text Corpora via Complex Deep Learning Naural Architectures

We analyze and compare five deep-learning neural architectures to manage the problem of irony and sarcasm detection for the Italian language. We briefly analyze the model architectures to choose the best compromise between performances and complexity. The obtained results show the effectiveness of such systems to handle the problem by achieving 93\% of F1-Score in the best case. As a case study, we also illustrate a possible embedding of the neural systems in a cloud computing infrastructure to exploit the computational advantage of using such an approach in tackling big data.

Modeling Personality Effect in Trust Prediction

Trust among users in online social networks is a key factor in determining the amount of information that is perceived as reliable. Compared to the number of users in online social networks, user-specified trust relations are very sparse. This makes the pair-wise trust prediction a challenging task. Social studies have investigated trust and why people trust each other. The relation between trust and personality traits of people who established those relations, has been proved by social theories. In this work, we attempt to alleviate the effect of the sparsity of trust relations by extracting implicit information from the users, in particular, by focusing on users' personality traits and seeking a low-rank representation of users. We investigate the potential impact on the prediction of trust relations, by incorporating users' personality traits based on the Big Five factor personality model. We evaluate the impact of similarities of users' personality traits and the effect of each personality trait on pair-wise trust relations. Next, we formulate a new unsupervised trust prediction model based on tensor decomposition. Finally, we empirically evaluate this model using two real-world datasets. Our extensive experiments confirm the superior performance of our model compared to the state-of-the-art approaches.

Making Use of More Reviews Skillfully in Explaninable Recommendation Gerneration

There are a lot of reviews in the Internet, and existing explainable recommendation techniques use them. However, how to use reviews has not been so far adequately addressed. This paper proposes a new exploiting method of reviews in explainable recommendation generation. Our new method makes use of not only reviews written but also those referred to by users. This paper adopts two state-of-the-art explainable recommendation approaches and shows how to apply our method to them. Moreover, our method in this paper considers the possibility of making use of reviews which do not provide detailed review utilization. Our proposal can be applied to different explainable recommendation approaches, which is shown by adopting the two approaches, with reviews that do not necessarily provide their detailed utilization data. The evaluation with using Amazon reviews shows an improvement of the two explainable recommendation approaches. Our proposal is the first attempt to make use of reviews which are written or referred to by users in generating explainable recommendation. Particularly, this study does not suppose that reviews provide their detailed utilization data.

A Penalty-Aware Cloud Monitoring System based on Blockchains

Today, traded cloud services are described by service level agreements that specify the obligations of providers such as availability or reliability. Violations of service level agreements lead to penalty payments. The recent development of prominent cloud platforms such as the re-design of Amazon's spot marketspace underpins a trend towards dynamic cloud markets where consumers migrate their services continuously to different marketspaces and providers to reach a cost-optimum. This leads to a heterogeneous IT infrastructure and consequently aggravates the monitoring of the delivered service quality. Hence, there is a need for a transparent penalty management system, which ensures that consumers automatically get penalty payments from providers in case of service violations. \newline In the paper at hand, we present a cloud monitoring system that is able to execute penalty payments autonomously. In this regard, we apply smart contracts hosted on blockchains, which continuously monitor cloud services and trigger penalty payments to consumers in case of service violations. For justification and evaluation we implement our approach by the IBM Hyperledger Fabric framework and create a use case with Amazon's cloud services as well as Azures cloud services to illustrate the universal design of the presented mechanism.

Bounded Graph Pattern Matching With View

Bounded evaluation using views is to compute the answers $Q({\cal D})$ to a query $Q$ in a dataset ${\cal D}$ by accessing only cached views and a small fraction $D_Q$ of ${\cal D}$ such that the size $|D_Q|$ of $D_Q$ and the time to identify $D_Q$ are independent of $|{\cal D}|$, no matter how big ${\cal D}$ is. Though proven effective for relational data, it has yet been investigated for graph data. In light of this, we study the problem of bounded pattern matching using views. We first introduce access schema ${\cal C}$ for graphs and propose a notion of joint containment to characterize bounded pattern matching using views. We show that a pattern query $\sq$ can be boundedly evaluated using views ${\cal V}(G)$ and a fraction $G_Q$ of $G$ if and only if the query $\sq$ is jointly contained by ${\cal V}$ and ${\cal C}$. Based on the characterization, we develop an efficient algorithm as well as an optimization strategy to compute matches by using ${\cal V}(G)$ and $G_Q$. Using real-life and synthetic data, we experimentally verify the performance of these algorithms, and show that (a) our algorithm for joint containment determination is not only effective but also efficient; and (b) our matching algorithm significantly outperforms its counterpart, and the optimization technique can further improve performance by eliminating unnecessary input.

Discover Relations in the Industry 4.0 Standards Via Unsupervised Learning on Knowledge Graph Embeddings

Industry 4.0 (I4.0) standards and standardization frameworks provide a unified way to describe smart factories. Standards specify the main components, systems, and processes inside a smart factory and the interaction among all of them. Furthermore, standardization frameworks classify standards according to their functions into layers and dimensions. Albeit informative, frameworks can categorize similar standards differently. As a result, interoperability conflicts are generated whenever smart factories are described with miss-classified standards. Approaches like ontologies and knowledge graphs enable the integration of standards and frameworks in a structured way. They also encode the meaning of the standards, known relations among them, as well as their classification according to existing frameworks. This structured modeling of the I4.0 landscape using a graph data model provides the basis for graph-based analytical methods to uncover alignments among standards. This paper contributes to analyzing the relatedness among standards and frameworks; it presents an unsupervised approach for discovering links among standards. The proposed method resorts to knowledge graph embeddings to determine relatedness among standards-based on similarity metrics. The proposed method is agnostic to the technique followed to create the embeddings and to the similarity measure. Building on the similarity values, community detection algorithms can automatically create communities of highly similar standards. Our approach follows the homophily principle, and assumes that related standards are together in a community. Thus, alignments across standards are predicted and interoperability issues across them are solved. We empirically evaluate our approach on a knowledge graph of 249 I4.0 standards using the Trans$^*$ family of embedding models for knowledge graph entities. Our results are promising and suggest that relations among standards can be detected accurately.

Sustainable Development Goal Relational Modelling and Prediction

A methodology for UN Sustainable Development Goal (SDG) attainment prediction is presented, the Sustainable Development Goals Correlation Attainment Predictions Extended framework SDG-CAP-EXT. Unlike previous SDG attainment methodologies, SDG-CAP-EXT takes into account the potential for a causal relationship between SDG indicators both with respect to the geographic entity under consideration (intra-entity) and neighbouring geographic entities to the current entity (inter-entity). The challenge is in the discovery of such causal relationships. A ensemble approach is presented that combines the results of a number of alternative causality relationship identification mechanisms. The identified relationships are used to build multi-variate time series prediction models that feed into a bottom-up SDG prediction taxonomy, which is used to make SDG attainment predictions and rank countries using a proposed Attainment Likelihood Index that reflects the likelihood of goal attainment. The framework is fully described and evaluated. The evaluation demonstrates that the SDG-CAP-EXT framework can produce better predictions than alternative models that do not consider the potential for intra- and inter-causal relationships.

An Approach to Develop Mobile Proxemic Applications

Traditional Human-Computer Interaction (HCI) is being overpowered by the widespread diffusion of smart and mobile devices. Currently, smart environments involve daily day activities covered by a huge variety of applications, which demand new HCI approaches. In this context, proxemic interaction, derived from the proxemic theory, becomes an influential approach to implement new kind of Mobile Human-Computer Interaction (MobileHCI) in smart environments. It is based on five proxemic dimensions: Distance, Identity, Location, Movement, and Orientation (DILMO). However, there is a lack of general and flexible tools and utilities focused on supporting the development of mobile proxemic applications. To respond to this need, we have previously proposed a framework for the design and implementation of proxemic applications for smart environments, whose devices interactions are defined in terms of DILMO dimensions. In this work, we extend this framework by integrating a Domain Specif Language (DSL) to support the designing phase. The framework also provides an API, that allows developers to simplify the process of proxemic information sensing (i.e., detection of DILMO dimensions) with mobile phones and wearable sensors. We perform an exhaustive revision of relevant and recent studies and describe in detail all components of our framework.

A Data Scope Management Service to Support Privacy by Design and GDPR Compliance

In order to empower user data protection and user rights, the European General Data Protection Regulation (GDPR) has been enforced. On the positive side, the user is obtaining advantages from GDPR. However, organisations are facing many difficulties in interpreting GDPR, and to properly applying it, and, in the meanwhile, due to their lack of compliance, many organisations are receiving huge fines from authorities. An important challenge is compliance with the Privacy by Design and by default (PbD) principles, which require that data protection is integrated into processing activities and business practices from the design stage. Recently, the European Data Protection Board (EDPB) released an official document with PbD guidelines, and there are various efforts to provide approaches to support these. However, organizations are still facing difficulties in identifying a flow for executing, in a coherent, linear and effective way, these activities, and a complete toolkit for supporting this. In this paper, we propose the design of such flow, and our comprehensive supporting toolkit, as part of the DEFeND EU Project platform. Within DEFeND, we identified candidate tools, fulfilling specific GDPR aspects, and integrated them in a comprehensive toolkit: the DEFeND Data Scope Management service (DSM). The aim of DSM is to support organizations for continuous GDPR compliance through model-based Privacy by Design analysis. Here, we present DSM, its design, flow, and a preliminary case study and evaluation performed with pilots from the healthcare, banking, public administration and energy sectors.