Search

Multi-task learning is becoming more popular and is being applied in a variety of applications. It improves the accuracy of prediction by simultaneously learning related tasks and saves cost through shared structures. In particular, the prediction of event type from social data is also an area where multi-task learning can be utilized. In this paper, we present a novel deep learning framework called S patial E vents P rediction (SEP) based on multi-task learning to predict the types of events that happen at a specific location from social data. The proposed model focuses on predicting the attribute types of an event, which is referred to as subtypes. Specifically, an event type-specific attention mechanism is introduced to extract the representations of social data and to identify their important components. The proposed attention mechanism is based on a two-level attention, which measures the importance of words and sentences to the subtypes of an event. We also propose a representation sharing method using semantic and spatial relationships between locations to alleviate the sparsity and incompleteness of data. The proposed representation sharing preserves the spatial heterogeneity between locations and significantly improves the accuracy of the overall framework. Experiments with real-world datasets confirm the effectiveness and efficiency of the proposed method.

There has been a great deal of interest in designing soft robots that can mimic a human system with haptic and proprioceptive functions. There is now a strong demand for soft robots that can sense their surroundings and functions in harsh environments. This is because the wireless sensing and actuating capabilities of these soft robots are very important for monitoring explosive gases in disaster areas and for moving through contaminated environments. To develop these wireless systems, complex electronic circuits must be integrated with various sensors and actuators. However, the conventional electronic circuits based on silicon are rigid and fragile, which can limit their reliable integration with soft robots for achieving continuous locomotion. In our study, we developed an untethered, soft robotic hand that mimics human fingers. The soft robotic fingers are composed of a thermally responsive elastomer composite that includes capsules of ethanol and liquid metals for its shape deformation through an electrothermal phase transition. And these soft actuators are integrated fully with flexible forms of heaters, with pressure, temperature, and hydrogen gas sensors, and wireless electronic circuits. Entire functions of this soft hand, including the gripping motion of soft robotic fingers and the real-time detections of tactile pressures, temperatures, and hydrogen gas concentrations, are monitored or controlled wirelessly using a smartphone. This wireless sensing and actuating system for somatosensory and respiratory functions of a soft robot provides a promising strategy for next-generation robotics.

Wireless communication through linkage with a smartphone and other portable devices in the sensor area are essential for increasing the efficiency of utilization by storing sensing-value data. Thus, the demand for wireless technology is increasing due to the advantages it provides for the various applications that use these data. However, there is still considerable ambiguity concerning the low portability of such technology due to the increased volume with component integration, the high consumption of power, and the necessity of having a separate battery. Herein, we present solutions for these problems with demonstrations that involve 1) the miniaturization of the device by altering the structure of the built-in battery, 2) the use of a pressure-activated switch for the low-power driving technology, and 3) the implementation of a wireless communication platform by integrating a Bluetooth module with the devices. In addition, we demonstrate a human-interactive display that enables users to instantly observe the changes in the brightness of the organic light-emitting diodes (OLED) as the pressure changes. We show that the system can display the measured, real-time pressure values on the screens of mobile devices, which represents a significant advancement in the fields of energy science and biomedical science.

News recommendation systems? purpose is to tackle the immense amount of news and offer personalized recommendations to users. A major issue in news recommendation is to capture the precise news representations for the efficacy of recommended items. Commonly, news contents are filled with well-known entities of different types. However, existing recommendation systems overlook exploiting external knowledge about entities and topical relatedness among the news. To cope with the above problem, in this paper, we propose Topic-Enriched Knowledge Graph Recommendation System(TEKGR). Three encoders in TEKGR handle news titles in two perspectives to obtain news representation embedding: (1) to extract meaning of news words without considering latent knowledge features in the news and (2) to extract semantic knowledge of news through topic information and contextual information from a knowledge graph. After obtaining news representation vectors, an attention network compares clicked news to the candidate news in order to get the user's final embedding. Our TEKGR model is superior to existing news recommendation methods by manipulating topical relations among entities and contextual features of entities. Experimental results on two public datasets show that our approach outperforms state-of-the-art deep recommendation approaches.

This paper presents an efficient method of extracting n-ary relations from multiple sentences which is called Entity-path and Discourse relation-centric Relation Extractor (EDCRE). Unlike previous approaches, the proposed method focuses on an entity link, which consists of dependency edges between entities, and discourse relations between sentences. Specifically, the proposed model consists of two main sub-models. The first one encodes sentences with a higher weight on the entity link while considering the other edges with an attention mechanism. To consider various latent discourse relations between sentences, the second sub-model encodes discourse relations between adjacent sentences considering the contents of each sentence. Experiment results on the cross-sentence relation extraction dataset, PubMed, and the document-level relation extraction dataset, DocRED, show that the proposed model outperforms state-of-the-art methods of extracting relations across sentences. Furthermore, ablation study proves that both the two main sub-models have noticeable effect on the relation extraction task.

A wearable smart sensing system is capable of continuously monitoring biometric information such as respiration, pulse, and body temperature while attached to an arbitrary surface on the user's body to be utilized freely during activities. Research conducted on new materials, fabrication processes, structure of electrodes, and wireless communication technologies has made constant progress in the development of smart sensing systems that use entirely wearable forms of devices to go beyond conventional devices that are based on intrinsically rigid components, such as silicon. Among various platforms that can be used in the development of smart sensing systems, optoelectronics provides innovative sensing functions (e.g., human–machine interfaces and phototherapy) that can be transferred from traditional healthcare to smart healthcare, such as point of care. Optoelectronics are light‐mediated displays that allow a light source to be controlled and a large light spectrum to be detected. Recently, this platform that uses smart and wearable optoelectronic sensing systems has become increasingly advanced to better help human beings treat their diseases through self‐healthcare. Herein, recent advanced technologies in materials, structures, and wireless platforms for smart sensors using wearable optoelectronics are reviewed.

Context-aware recommender systems (CARS) exploit multiple contexts to improve user experience in embracing new information and services. Tensor factorization (TF), a type of latent factor model, has achieved remarkable performance in CARS. TF learns latent representations of contexts by decomposing an observed rating tensor and combines the latent representations as a vector form to represent contextual influence on users and items. However, due to the limitation of the contextual expression power, they have difficulties in effectively capturing complex correlations among multiple contexts, and also the meaning of each context is diluted. To address the issue, we propose a reliable TF-based recommender system based on a proposed context tensor (CT-CARS), which incorporates a variety of correlations among contexts. CT-CARS contains a novel recommendation rating function and a learning algorithm. Specifically, the proposed context tensor elaborately captures the influences of both individual contexts and context combinations. Moreover, we introduce a novel parameter initialization based on past-learned results to improve the reliability of recommendations. The overall time complexity of our parameter learning algorithm grows linearly as dataset size increases. Experiments on six real-world datasets including two large-scaled datasets show that CT-CARS outperforms the existing state-of-the-art models in terms of both accuracy and reliability.

The accurate prediction of stress in a person’s life has a significant effect on improving personal health and the national economy. Since individuals have different historical circumstances and personality traits, stress symptoms and levels may vary from person to person. Thus, most studies on stress prediction pay attention to personalized models, which determine the personal stress level using user-specific information and heterogeneous stress-related data. However, these models cannot elaborately handle the uncertainty caused by the sparsity, data imbalance, irregularity, and high-dimensionality of user-specific information. In particular, out-of-sample users increase uncertainty. To cope with the problem, we propose a personalized stress-level prediction model with inductive Gaussian representation (PSP-IGR), which exploits heterogeneous inputs with a unified end-to-end approach. PSP-IGR extracts feature vectors from the heterogeneous inputs via Gaussian sampling, domain rules, and deep learning, depending on the characteristics of each input. Especially, PSP-IGR inductively generates a Gaussian feature vector called IGR by Gaussian sampling from the shared contents of user-specific information. Thus, PSP-IGR not only generalizes to both in-sample and out-of-sample users effectively but also deals with the uncertainty problem caused by limitations of healthcare datasets. Also, since we fuse the extracted feature vectors considering their characteristics (Gaussian and point vectors), we can preserve the expressiveness of each feature vector. Experiments on a real-world dataset, including survey results, wearable sensor signals, and contexts, demonstrate that PSP-IGR shows higher accuracy in predicting individual stress-level than previous models.

A large number of spatial knowledge graphs (SKGs) are available from spatially enriched knowledge bases, e.g., DBpedia and YAGO2. This provides a great chance to understand valuable information about the regions surrounding us. However, it is hard to comprehend SKGs due to the explosively growing volume and the complication of the graph structures. Thus we study the problem of similar region search (SRS), which is an easy-to-use but effective way to explore spatial data. The effectiveness of SRS highly depends on how to measure the region similarity. However, existing approaches cannot make use of the rich information contained in SKGs thus may lead to incorrect results. In this paper, we propose a spatial knowledge representation learning method for region similarity, namely SKRL4RS. SKRL4RS firstly encodes the spatial entities of an SKG into a vector space to make it easier to extract useful features. Then regions are represented by 3-D tensors using the spatial entity embeddings together with geographical information. Finally, region tensors are fed into the conventional triplet network to learn the feature vectors of regions. The region similarity measure learned by SKRL4RS can capture the hierarchical types, semantic relatedness, and relative locations of spatial entities inside a region. Experimental results on two real-world datasets show that our SKRL4RS outperforms the state-of-the-art by a significant margin in terms of the accuracy of measuring region similarity.

To share and publish the domain knowledge of IoT objects, the development of a semantic IoT model based directory system that manages meta-data and relationships of IoT objects is required. Many researches focus on static relationships between IoT objects. However, because complex relationships between various resources change with time in an IoT environment, an efficient method for updating the meta-data is required. Thus, we propose an IoT-DS as the IoT directory that supports semantic description, discovery, and integration of IoT objects. Firstly, we introduce a semantic IoT component model to establish a shared conceptualization. Secondly, we present general cases of relationships to efficiently interact between IoT-DS and IoT objects. Thirdly, we construct IoT-DS as a Web portal. Finally, we verify in our evaluation study that the query processing time and communication workload imposed by the proposed approach are reduced.

In real world, a service may be affected by various factors making service selection process a difficult task. Therefore it is important to select an appropriate service efficiently from a large number of services based on the quality of a service QoS. However, existing approaches have limitations in processing QoS data with outliers that may occur in reality. When decomposing a global QoS constraint given by a service requester into local constraints, local thresholds are commonly computed with no consideration of outliers. Therefore, an appropriate service may not be selected due to false negative errors. To address this limitation, we propose an outlier-robust Web service selection approach based on a probabilistic QoS model. Specifically, the approach prunes services which do not satisfy a global constraint due to outliers. A service is then selected based on its probability of satisfying a global constraint. Experimental results show that the proposed approach outperforms an existing solution in terms of global QoS conformance.