An e-TextBook can serve as an interactive learning environment (ILE), facilitating more effective teaching and learning processes. In this paper, we propose the novel concept of an EPUB 3-based Hybrid e-TextBook, which allows for interaction between the digital and the physical world. In that regard, we first investigated the gap between the expectations of teachers with respect to e-TextBook functionalities, on the one hand, and the ILE functionalities offered by e-TextBooks, on the other hand. Next, together with teachers, we co-designed and developed prototype EPUB 3-based Hybrid e-TextBooks that make it possible to connect their learning content to smart devices in classrooms, leveraging both digital publishing and Semantic Web tools. Based on experimentation with our prototype Hybrid e-TextBooks, we can argue that a semantically enriched EPUB 3-based Hybrid e-TextBook is able to act as a comprehensive ILE, providing the tools needed by teachers in smart classrooms. Furthermore, expert observations and Smiley o'meter results demonstrate an effective impact on student cognition and motivation.

Currently, temperature-based condition monitoring cannot be used to accurately identify potential faults early in a rotating machines’ lifetime since temperature changes are only detectable when the fault escalates. However, currently only point measurements, i.e. thermocouples, are used. In this article, infrared thermal imaging is used which – as opposed to simple thermocouples – provides spatial temperature information. This information proves crucial for the identification of several machine conditions and faults. In this paper the conditions considered are outer-raceway damage in bearings, hard-particle contamination in lubricant and several gradations of shaft imbalance. The fault detection is done using an image processing and machine learning solution which can accurately detect the majority of the faults and conditions in our data set.

In the Internet of Things (IoT), data-producing entities sense their environment and transmit these observations to a data processing platform for further analysis. Applications can have a notion of context awareness by combining this sensed data, or by processing the combined data. The processes of combining data can consist both of merging the dynamic sensed data, as well as fusing the sensed data with background and historical data. Semantics can aid in this task, as they have proven their use in data integration, knowledge exchange and reasoning. Semantic services performing reasoning on the integrated sensed data, combined with background knowledge, such as profile data, allow extracting useful information and support intelligent decision making. However, advanced reasoning on the combination of this sensed data and background knowledge is still hard to achieve. Furthermore, the collaboration between semantic services allows to reach complex decisions. The dynamic composition of such collaborative workflows that can adapt to the current context, has not received much attention yet. In this paper, we present MASSIF, a data-driven platform for the semantic annotation of and reasoning on IoT data. It allows the integration of multiple modular reasoning services that can collaborate in a flexible manner to facilitate complex decision-making processes. Data-driven workflows are enabled by letting services specify the data they would like to consume. After thorough processing, these services can decide to share their decisions with other consumers. By defining the data these services would like to consume, they can operate on a subset of data, improving reasoning efficiency. Furthermore, each of these services can integrate the consumed data with background knowledge in its own context model, for rapid intelligent decision making. To show the strengths of the platform, two use cases are detailed and thoroughly evaluated.

The condition of a machine can automatically be identified by creating and classifying features that summarize characteristics of measured signals. Currently, experts, in their respective fields, devise these features based on their knowledge. Hence, the performance and usefulness depends on the expert's knowledge of the underlying physics or statistics. Furthermore, if new and additional conditions should be detectable, experts have to implement new feature extraction methods. To mitigate the drawbacks of feature engineering, a method from the subfield of feature learning, i.e., deep learning (DL), more specifically convolutional neural networks (NNs), is researched in this paper. The objective of this paper is to investigate if and how DL can be applied to infrared thermal (IRT) video to automatically determine the condition of the machine. By applying this method on IRT data in two use cases, i.e., machinefault detection and oil-level prediction, we show that the proposed system is able to detect many conditions in rotating machinery very accurately (i.e., 95 and 91.67% accuracy for the respective use cases), without requiring any detailed knowledge about the underlying physics, and thus having the potential to significantly simplify condition monitoring using complex sensor data. Furthermore, we show that by using the trained NNs, important regions in the IRT images can be identified related to specific conditions, which can potentially lead to new physical insights.

This paper presents the concept and the realisation of an interactive multimedia installation, called BilliArT, together with an explorative user study conducted on the data gathered during a public exhibition of the installation. The study concerns functional properties of the installation (e.g. usability, design quality) and subjective qualities of the sonic and visual feedback of the installation. The installation consists in a collaborative environment based on the carambole billiards game, which allows the users to engage in a user-driven machine-based jazz-inspired music improvisation, augmented with visual feedback. The installation is designed to promote the interaction among the users and the billiard game, stimulating the motivation to engage in the game by balancing predictable and unpredictable output, and reinforcing the feeling of reward, irrespective of their level of musical training. BilliArT introduces a new framework for expressive interaction related to the concepts of motivation and reward. The exploratory study proved the ability of the installation to activate the users' sense of aesthetic reward, leading to a more active and satisfactory engagement in the game. Future studies may exploit these results to the advantage of the world of the arts, as well as of studies in human-computer interaction, interface design, and cultural heritage preservation.

The hashtag is a recognized and often used method to collect Twitter messages. However, it has its limits with respect to the inclusion of follow-messages, or @replies, that do not contain a hashtag. This paper explored to what extent the inclusion of non-hashtagged responses affected the study of interactions between Twitter users. We drew from the Twitter debate on the 2014 Belgian elections, collected under the #vk2014 hashtag. Our dataset included non-hashtagged responses to assess (1) how they differ from hashtagged responses; and, (2) how this affects the conversation network. The findings showed that (1) hashtagged responses were more likely to include other interactive elements (e.g., hyperlinks); and, (2) the inclusion of non-hashtagged responses generated larger and more reciprocal networks. However, central users further strengthened their position in the network.

This paper presents a novel wavelet-based transform and coding scheme for irregular meshes. The transform preserves geometric features at lower resolutions by adaptive vertex sampling and retriangulation, resulting in more accurate subsampling and better avoidance of smoothing and aliasing artefacts. By employing octree-based coding techniques, the encoding of both connectivity and geometry information is decoupled from any mesh traversal order, and allows for exploiting the intra-band statistical dependencies between wavelet coefficients. Improvements over the state of the art obtained by our approach are three-fold: 1 improved rate-distortion performance over Wavemesh and IPR for both the Hausdorff and root mean square distances at low-to-mid-range bitrates, most obvious when clear geometric features are present while remaining competitive for smooth, feature-poor models; 2 improved rendering performance at any triangle budget, translating to a better quality for the same runtime memory footprint; 3 improved visual quality when applying similar limits to the bitrate or triangle budget, showing more pronounced improvements than rate-distortion curves.

Vibration analysis is a well-established technique for condition monitoring of rotating machines as the vibration patterns differ depending on the fault or machine condition. Currently, mainly manually-engineered features, such as the ball pass frequencies of the raceway, RMS, kurtosis an crest, are used for automatic fault detection. Unfortunately, engineering and interpreting such features requires a significant level of human expertise. To enable non-experts in vibration analysis to perform condition monitoring, the overhead of feature engineering for specific faults needs to be reduced as much as possible. Therefore, in this article we propose a feature learning model for condition monitoring based on convolutional neural networks. The goal of this approach is to autonomously learn useful features for bearing fault detection from the data itself. Several types of bearing faults such as outer-raceway faults and lubrication degradation are considered, but also healthy bearings and rotor imbalance are included. For each condition, several bearings are tested to ensure generalization of the fault-detection system. Furthermore, the feature-learning based approach is compared to a feature-engineering based approach using the same data to objectively quantify their performance. The results indicate that the feature-learning system, based on convolutional neural networks, significantly outperforms the classical feature-engineering based approach which uses manually engineered features and a random forest classifier. The former achieves an accuracy of 93.61 percent and the latter an accuracy of 87.25 percent.

Linked Data storage solutions often optimize for low latency querying and quick responsiveness. Meanwhile, in the back-end, offline ETL processes take care of integrating and preparing the data. In this paper we explain a workflow and the results of a benchmark that examines which Linked Data storage solution and setup should be chosen for different dataset sizes to optimize the cost-effectiveness of the entire ETL process. The benchmark executes diversified stress tests on the storage solutions. The results include an in-depth analysis of four mature Linked Data solutions with commercial support and full SPARQL 1.1 compliance. Whereas traditional benchmarks studies generally deploy the triple stores on premises using high-end hardware, this benchmark uses publicly available cloud machine images for reproducibility and runs on commodity hardware. All stores are tested using their default configuration. In this setting Virtuoso shows the best performance in general. The other tree stores show competitive results and have disjunct areas of excellence. Finally, it is shown that each store's performance heavily depends on the structural properties of the queries, giving an indication of where vendors can focus their optimization efforts.

Linked Data interfaces exist in many flavours, as evidenced by subject pages, sparql endpoints, triple pattern interfaces, and data dumps. These interfaces are mostly used to retrieve parts of a complete dataset, such parts can for example be defined by ranges in one or more dimensions. Filtering Linked Data by dimensions such as time range, geospatial area, or genomic location, requires the lookup of data within ordinal ranges. To make retrieval by such ranges generic and cost-efficient, we propose a rest solution in-between looking up data within ordinal ranges entirely on the server, or entirely on the client. To this end, we introduce a method for extending any Linked Data interface with an n-dimensional interface-level index such that n-dimensional ordinal data can be selected using n-dimensional ranges. We formally define Range Gates and Range Fragments and theoretically evaluate the cost-efficiency of hosting such an interface. By adding a multidimensional index to a Linked Data interface for multidimensional ordinal data, we found that we can get benefits from both worlds: the expressivity of the server raises, yet remains more cost-efficient than an interface providing the full functionality on the server-side. Furthermore, the client now shares in the effort to filter the data. This makes query processing becomes more flexible to the end-user, because the query plan can be altered by the engine. In future work we hope to apply Range Gates and Range Fragments to real-world interfaces to give quicker access to data within ordinal ranges.

Algorithmic storytelling over Linked Data on the Web is a challenging task in which many graph-based pathfinding approaches experience issues with consistency regarding the resulting path that leads to a story. In order to mitigate arbitrariness and increase consistency, we propose to improve the semantic relatedness of concepts mentioned in a story by increasing the relevance of links between nodes through additional domain delineation and refinement steps. On top of this, we propose the implementation of an optimized algorithm controlling the pathfinding process to obtain more homogeneous search domain and retrieve more links between adjacent hops in each path. Preliminary results indicate the potential of the proposal.

In order to reduce the server-side cost of publishing queryable Linked Data, Triple Pattern Fragments (TPF) were introduced as a simple interface to RDF triples. They allow for SPARQL query execution at low server cost, by partially shifting the load from servers to clients. The previously proposed query execution algorithm provides a solution that is highly inefficient, often requiring an amount of HTTP calls that is magnitudes larger than the optimal solution. We have proposed a new query execution algorithm with the aim to solve this problem. Our solution significantly improves on the current work by maintaining a complete overview of the query instead of just looking at local optima. In this paper, we describe a demo that allows a user to easily compare the results of both implementations. We show both query results and number of executed HTTP calls, proving a clear picture of the difference between the two algorithms.

Routing and navigation services for leisure activities are conditioned by special needs and trade-offs. The advent of online communities and large crowdsourced datasets offers opportunities to improve the adoption of a user's perspective in these suggested paths. This paper focuses on achieving two goals. First, the presented methodology analyses a dataset of 190,610 historical GPS traces to gain insights into the appreciation or attractiveness of each edge in a real-world network for a specific leisure activity (i.e. road cycling). Second, as literature on these leisure activities is still sparse, we want to create a thorough understanding of the activities at hand for future work. An appreciation model is proposed and the spread of this score is analyzed in shortest-path alternatives of popular routing engines for this activity. This analysis successfully discriminates these shortest paths based on the scoring value and three morphological parameters of the path. However, the robustness of the model should be improved to ensure the viability of the proposed approach in future work. More specifically, further research on the local optimality of the route choices will be imperative.

Provenance and other metadata are essential for determining ownership and trust. Nevertheless, no systematic approaches were introduced so far in the Linked Data publishing workow to capture them. Dening such metadata remained independent of the rdf data generation and publishing. In most cases, metadata is manually dened by the data publishers (person-agents), rather than produced by the involved applications (software-agents). Moreover, the generated rdf data and the published one are considered to be one and the same, which is not always the case, leading to pure, condense and often seductive information. This paper introduces an approach that relies on declarative descriptions of (i) mapping rules, specifying how the rdf data is generated, and of (ii) raw data access interfaces to automatically and incrementally generate provenance and metadata information. This way, it is assured that the metadata information is accurate, consistent and complete.

This paper sheds light on the different interaction types among social media users that benefit information diffusion and provenance analysis. In particular, we identify explicit and implicit interactions in Twitter, including informal conventions applied by users. In our empirical evaluation considering only retweets, the most common means of information propagation in Twitter, we can infer 50% of message provenance. However, if we consider other types of interactions, we can explain another 13%. Accordingly, we enrich the PROV-SAID model for information diffusion, which extends the W3C PROV standard for provenance.