Use Of Contextual Information Research Articles

Densely connected network with improved pyramidal bottleneck residual units for super-resolution

Recent studies have shown that super-resolution can be significantly improved by using deep convolution neural network. Although applying a larger number of convolution kernels can extract more features, increasing the number of feature mappings will dramatically increase the training parameters and time complexity. In order to balance the workload among all units and maintain appropriate time complexity, this paper proposes a new network structure for super-resolution. For the sake of making full use of context information, in the structure, the operations of division (S) and fusion (C) are added to the pyramidal bottleneck residual units, and the dense connected methods are used. The proposed network include a preliminary feature extraction net, seven residual units with dense connections, seven convolution layers with the size of 1×1 after each residual unit, and a deconvolution layer. The experimental results show that the proposed network has better performance than most existing methods.

MCnet: Multiple Context Information Segmentation Network of No-Service Rail Surface Defects

Surface defect segmentation of no-service rail is important for its quality assessment. There are several challenges of uneven illumination, complex background, and difficulty of sample collection for no-service rail surface defects (NRSDs). In this article, we propose an acquisition scheme with two lamp light and color scan line charge-coupled device (CCD) to alleviate uneven illumination. Then, a multiple context information segmentation network is proposed to improve NRSD segmentation. The network makes full use of context information based on dense block, pyramid pooling module, and multi-information integration. Besides, the attention mechanism is applied to optimize extracted information by filtering noise. For the problem of real sample shortage, we propose to utilize artificial samples to train the network. And an NRSD data set NRSD-MN is built with artificial NRSDs and natural NRSDs. Experimental results show that our method is feasible and has a good segmentation effect on artificial and natural NRSDs.

Age-related dissociation of N400 effect and lexical priming

The use of contextual information is an important capability to facilitate language comprehension. This can be shown by studying behavioral and neurophysiological measures of accelerated word recognition when semantically or phonemically related information is provided in advance, resulting in accompanying attenuation of the respective event-related potential, i.e. the N400 effect. Against the background of age-dependent changes in a broad variety of lexical capacities, we aimed to study whether word priming is accomplished differently in elderly compared to young persons. 19 young (29.9 ± 5.6 years) and 15 older (69.0 ± 7.2 years) healthy adults participated in a primed lexical decision task that required the classification of target stimuli (words or pseudo-words) following related or unrelated prime words. We assessed reaction time, task accuracy and N400 responses. Acceleration of word recognition by semantic and phonemic priming was significant in both groups, but resulted in overall larger priming effects in the older participants. Compared with young adults, the older participants were slower and less accurate in responding to unrelated word-pairs. The expected N400 effect was smaller in older than young adults, particularly during phonemic word and pseudo-word priming, with a rather similar N400 amplitude reduction by semantic relatedness. The observed pattern of results is consistent with preserved or even enhanced lexical context sensitivity in older compared to young adults. This, however, appears to involve compensatory cognitive strategies with higher lexical processing costs during phonological processing in particular, suggested by a reduced N400 effect in the elderly.

Building extraction from remote sensing image based on improved segnet neural network and image pyramid

Aiming at the problems such as the insufficient use of context information in high-resolution remote sensing image by convolutional neural network and the time-consuming training model, a building extraction method based on the combination of improved Segnet neural network and image pyramid is proposed. First, the Gaussian and Laplacian pyramid algorithms are used to obtain the multi-scale and multi-resolution detailed feature images of the original image respectively, and the network is trained together with the original image, so that the model can learn and extract more abundant image features. Secondly, in order to reduce the complexity of the network, reduce the encoder and decoder structure in the original Segnet neural network, save the calculation cost and effectively avoid the occurrence of over-fitting and other phenomena. The experiment uses Massachusetts buildings as the dataset. Compared with the original segnet neural network extraction method, the proposed method has significantly improved the extraction accuracy and extraction efficiency.

AAMDRL: Augmented Asset Management With Deep Reinforcement Learning

Can an agent learn efficiently in a noisy and self adapting environment with sequential, non-stationary and non-homogeneous observations? Through trading bots, we illustrate how Deep Reinforcement Learning (DRL) can tackle this challenge. Our contributions are threefold: (i) the use of contextual information also referred to as augmented state in DRL, (ii) the impact of a one period lag between observations and actions that is more realistic for an asset management environment , (iii) the implementation of a new repetitive train test method called walk forward analysis, similar in spirit to cross validation for time series. Although our experiment is on trading bots, it can easily be translated to other bot environments that operate in sequential environment with regime changes and noisy data. Our experiment for an augmented asset manager interested in finding the best portfolio for hedging strategies shows that AAMDRL achieves superior returns and lower risk.

Fuzzy, Dynamic and Trust Based Routing Protocol for IoT

Internet of Things (IoT) could be described as the pervasive and global network where real-world entities augmented with computing devices, sensors and actuators are connected to the Internet, enabling them to publish their generated data. Thus, an efficient and secure routing service is required to enable efficient communication of information among IoT nodes. This sophisticated, dynamic, and ultra-large-scale network requires the use of contextual information, attention to security issues and the consideration of service quality to make proper routing decisions. The routing protocol for low-power and lossy networks (RPL) and improved versions of it are experiencing severe performance gaps under network attacks such as BLACKHOLE, SYBIL and RANK. This paper uses the concept of trust as an umbrella to cover countermeasures for addressing the consequences of attacks. Accordingly, a multi-fuzzy, dynamic and hierarchical trust model (FDTM-IoT) is proposed. The main dimensions of this model are contextual information (CI), quality of service (QoS) and quality of P2P communication (QPC). Each dimension also has its own sub-dimensions or criteria. FDTM-IoT is integrated into RPL (FDTM-RPL) as objective function. FDTM-RPL use trust to deal with attacks. In the proposed method, fuzzy logic has been used in trust calculations to consider uncertainty as one of the most important inherent characteristics of trust. The efficiency of FDTM-RPL in various scenarios (including small-scale to large-scale networks, mobile environment as well as different transmission rates and under different attacks) has been compared with standard RPL protocols. FDTM-RPL provides high performance in detecting attacks. Additionally, it improves network performance in a variety of criteria, including end-to-end delay and packet loss rates.

Graph LSTM with Context-Gated Mechanism for Spoken Language Understanding

Much research in recent years has focused on spoken language understanding (SLU), which usually involves two tasks: intent detection and slot filling. Since Yao et al.(2013), almost all SLU systems are RNN-based, which have been shown to suffer various limitations due to their sequential nature. In this paper, we propose to tackle this task with Graph LSTM, which first converts text into a graph and then utilizes the message passing mechanism to learn the node representation. Not only the Graph LSTM addresses the limitations of sequential models, but it can also help to utilize the semantic correlation between slot and intent. We further propose a context-gated mechanism to make better use of context information for slot filling. Our extensive evaluation shows that the proposed model outperforms the state-of-the-art results by a large margin.

Efficient Melody Extraction Based on Extreme Learning Machine

Melody extraction is an important task in music information retrieval community and it is unresolved due to the complex nature of real-world recordings. In this paper, the melody extraction problem is addressed in the extreme learning machine (ELM) framework. More specifically, the input musical signal is first pre-processed to mimic the human auditory system. The music features are then constructed by constant-Q transform (CQT), and the concentration strategy is introduced to make use of contextual information. Afterwards, the rough melody pitches are determined by ELM network, according to its pre-trained parameters. Finally, the rough melody pitches are fine-tuned by the spectral peaks around the frame-wise rough pitches. The proposed method can extract melody from polyphonic music efficiently and effectively, where pitch estimation and voicing detection are conducted jointly. Some experiments have been conducted based on three publicly available datasets. The experimental results reveal that the proposed method achieves higher overall accuracies with very fast speed.

Relationship-Embedded Representation Learning for Grounding Referring Expressions.

Grounding referring expressions in images aims to locate the object instance in an image described by a referring expression. It involves a joint understanding of natural language and image content, and is essential for a range of visual tasks related to human-computer interaction. As a language-to-vision matching task, the core of this problem is to not only extract all the necessary information (i.e., objects and the relationships among them) in both the image and referring expression, but also make full use of context information to align cross-modal semantic concepts in the extracted information. Unfortunately, existing work on grounding referring expressions fails to accurately extract multi-order relationships from the referring expression and associate them with the objects and their related contexts in the image. In this paper, we propose a cross-modal relationship extractor (CMRE) to adaptively highlight objects and relationships (spatial and semantic relations) related to the given expression with a cross-modal attention mechanism, and represent the extracted information as a language-guided visual relation graph. In addition, we propose a Gated Graph Convolutional Network (GGCN) to compute multimodal semantic contexts by fusing information from different modes and propagating multimodal information in the structured relation graph. Experimental results on three common benchmark datasets show that our Cross-Modal Relationship Inference Network, which consists of CMRE and GGCN, significantly surpasses all existing state-of-the-art methods.

The intuitive use of contextual information in decisions made with verbal and numerical quantifiers

Verbal and numerical formats (e.g., verbal: “low fat,” or numerical: “20% fat”) are used interchangeably to communicate nutritional information. However, prior research implies that verbal quantifiers are processed more intuitively than numerical ones. We tested this hypothesis in two pre-registered experiments measuring four indicators of processing style: (a) response time, (b) decision performance, (c) reliance on irrelevant contextual information, which we inferred from participants’ decision patterns, and (d) the level of interference from a concurrent memory task. Participants imagined they had consumed a given amount of a nutrient (represented in a pie chart) and decided whether a new quantity (either verbal or numerical) could be eaten within their guideline daily amount (GDA). The experiments used a mixed design varying format (verbal or numerical), concurrent memory load (no load, easy, and hard load in Experiment 1; no load and hard load in Experiment 2), nutrient (fat and minerals), quantity (low, medium, and high in Experiment 1; low and high in Experiment 2), and the assigned correct response for a trial (within and exceeding limits). Participants were faster and made fewer correct decisions with verbal quantifiers, and they relied more on contextual information (i.e., the identity of the nutrient involved). However, memory load did not impair decisions with verbal or numerical quantifiers. Altogether, these results suggest that verbal quantifiers are processed intuitively, slightly more so than numerical quantifiers, but that numerical quantifiers do not require much analytical processing to reach simple decisions.

AAFM: Adaptive Attention Fusion Mechanism for Crowd Counting

CNN-based crowd counting methods have achieved great progress in recent years. However, most of these CNN-based crowd counting methods do not make full use of contextual information, which contains high-level semantic features and low-level detail features from different receptive fields of CNN. But rich contextual information is important to solve the scale variation problem of crowd counting. So the precision of previous CNN-based crowd counting methods is decreased. To solve this problem, we propose an adaptive attention fusion mechanism (AAFM). AAFM can use multi-scale features from different receptive fields of CNN effectively. It integrates the convolution network for feature learning and the attention mechanism for multi-scale features fusion. We apply the first 13 convolution layers of VGG-16 as the encoder module to extract the base features. Then, the base features are fed into the decoder module. The decoder module mainly contains Density Regression Branch (DRB) and Feature Fusion Branch (FFB). DRB uses multiple convolution layers for feature learning and multi-scale feature extraction. FFB uses attention modules for modeling multi-scale features and element-wise multiply for features fusion. Therefore, AAFM can obtain rich contextual information into the encoder-decoder framework for generating high-quality crowd density maps and accurate counting. We perform experiments on ShanghaiTech, UCF-CC-50, and UCF-QNRF datasets, and AAFM achieves promising results.

A Text-Granulation Clustering Approach With Semantics for E-Commerce Intelligent Storage Allocation

With the hot rise and increasing popularity of e-commerce and online shopping, exponentially growing orders require to be picked up timely. In the order-based warehouse picking process, more intelligent optimization methods are urgently demanded to meet the maturity of robot technology. Due to the wide variety of SKUs (stock-keeping units) in warehouses, direct analysis based on orders can hardly get strong correlations among SKUs. Since orders and SKUs usually are presented in text form files, this paper proposes a clustering optimization algorithm based on text information granules which makes full use of context information and semantics. Storage assignment is determined by clustering SKUs first and then performing correlation analysis on the clustering results according to orders. The algorithm can give the semantic description of each cluster so that the clustering process is interpretable and transparent. Extensive experiments on a distribution center of a clothing production enterprise indicate 40.5 to 57.9% improvement in the average order picking distance compared with the ABC classification.

Learning Chinese Word Segmentation Based on Bidirectional GRU-CRF and CNN Network Model

Chinese word segmentation is the basis of the Chinese natural language processing (NLP). With the development of the deep learning, various neural network models are applied to the Chinese word segmentation. However, current neural network models have the characteristics of artificial feature extraction, nonstandard word-weight, inability to effectively use long-distance information and long training time of models in Chinese word segmentation. To solve a series of problems, this article presents a CNN-Bidirectional GRU-CRF neural network model (CNN Bidirectional GRU CRF Network, CBiGCN), which breaks through the limit of conventional method window, truly realizes end-to-end processing and applies to the neural network model by the five-Tag set method, bias-variable-weight greedy strategy and supplements by Goldstein-Armijo guidelines. Besides, this model, with simple structure, is easy to be operated. And it can automatically learn features, reduces large amounts of tasks on specific knowledge in the form of handcrafted features and data pre-processing, makes use of context information effectively. The authors set an experiment with two data corpuses for Chinese word segmentation to evaluate their system. The experiment verified their new model can obtain better Chinese word segmentation results and greatly reduce training time.

Web video classification with visual and contextual semantics

SummaryOn the social Web, the amount of video content either originated from wireless devices or previously received from media servers has increased enormously in the recent years. The astounding growth of Web videos has stimulated researchers to propose new strategies to organize them into their respective categories. Because of complex ontology and large variation in content and quality of Web videos, it is difficult to get sufficient, precisely labeled training data, which causes hindrance in automatic video classification. In this paper, we propose a novel content‐ and context‐based Web video classification framework by rendering external support through category discriminative terms (CDTs) and semantic relatedness measure (SRM). Mainly, a three‐step framework is proposed. Firstly, content‐based video classification is proposed, where twofold use of high‐level concept detectors is leveraged to classify Web videos. Initially, category classifiers induced from VIREO‐374 detectors are trained to classify Web videos, and then concept detectors with high confidence for each video are mapped to CDT through SRM‐assisted semantic content fusion function to further boost the category classifiers, which intuitively provide a more robust measure for Web video classification. Secondly, a context‐based video classification is proposed, where twofold use of contextual information is also harnessed. Initially, cosine similarity and then semantic similarity are measured between text features of each video and CDT through vector space model (VSM)‐ and SRM‐assisted semantic context fusion function, respectively. Finally, classification results from content and context are fused to compensate for the shortcomings of each other, which enhance the video classification performance. Experiments on large‐scale video dataset validate the effectiveness of the proposed solution.

CD-CARS: Cross-domain context-aware recommender systems

In this paper, we address two research topics in Recommender Systems (RSs) which have been developed in parallel without a deeper integration: Cross-Domain RS (CDRS) and Context-Aware RS (CARS). CDRS have emerged to enhance the quality of recommendations in a target domain by leveraging sources of information in different domains. CDRS are especially useful to address cold-start, sparsity and diversity problems in target domains with scarce information. CARS, on its turn, have been proposed to consider contextual information for recommendations. Such systems are suitable when the users’ interests change according to factors like time, location, among others. By combining these two approaches, better RSs can be developed, considering both the availability of useful data from multiple domains and the use of contextual information. In this paper, we formalize the combination of CDRS and CARS, which represents a more systematic integration of these approaches compared to previous work. Based on this formulation, we developed novel RSs techniques, named CD-CARS. To evaluate the developed CD-CARS techniques, we performed extensive experimentation through real datasets taking into account several scenarios. The recommendations were evaluated in terms of predictive and ranking performance, respectively achieving up to 62.6% and 45%, depending on the scenario, in comparison to traditional cross-domain collaborative filtering techniques. Therefore, the experimental results have shown that the integration of techniques developed in isolation can be useful in a variety of situations, in which recommendations can be improved by information gathered from different sources and can be refined by considering specific contextual information.

Modeling Mobile Edge Computing Deployments for Low Latency Multimedia Services

Multi-access edge computing (MEC) technologies bring important improvements in terms of network bandwidth, latency, and use of context information and critical for services like multimedia streaming, augmented, and virtual reality. In future deployments, operators will need to decide how many MEC points of presence (PoPs) are needed and where to deploy them, also considering the number of base stations needed to support the expected traffic. This paper presents an application of inhomogeneous Poisson point processes with hard-core repulsion to model feasible MEC infrastructure deployments. With the presented methodology a mobile network operator knows where to locate the MEC PoPs and associated base stations to support a given set of services. We evaluate our model with simulations in realistic scenarios, namely Madrid City Center, an industrial area and a rural area.

Object codetection based on a higher-order conditional random field

In recent years, codetecting objects through the use of contextual information across multiple images has attracted considerable attention. We introduce an object codetection method that exploits contextual information among multiple images through a higher-order conditional random field (CRF). First, we obtain object candidates from each image of a test set by using a pretrained detector. Second, we feed the object candidates into a higher-order CRF that captures the appearance similarity using pairwise potentials and object category co-occurrence constraints using higher-order potentials. Finally, we jointly predict the category labels of all object candidates through the mean field inference in the CRF. Experimental results on the Caltech Pedestrian, PASCAL VOC 2007, PASCAL VOC 2012, and COCO datasets demonstrate the effectiveness of the proposed method compared to the baseline method.

Robust visual tracking based on spatial context pyramid

In recent years, discriminative correlation filter (DCF) has gained a lot of popularity in visual tracking, mainly due to its circular sampling from limited training data and computational efficiency in Fourier domain. However, such trackers do not make reasonable use of context information, resulting in limited performance. In this paper, we propose a novel DCF tracking framework based on spatial context pyramid (SCPT) to overcome this problem. Firstly, we take global spatial context into account to exploit the relationship between the target and its context for better tracking. Secondly, we design an effective spatial window to highlight the target while suppressing the background, and thus a robust filter model which has a high response for the target and low response for the background can be learned. Thirdly, we construct a context pyramid representation using multi-level spatial windows for adapting different challenging factors. To validate the compatibility of the proposed algorithm, we implement two versions with the representations from both conventional features and deep convolutional neural network (CNN) features. Extensive experimental results on the OTB-2013 benchmark demonstrate the effectiveness of the proposed tracker in terms of accuracy and robustness.

The use of contextual information in forensic toxicology: An international survey of toxicologists' experiences

Cognitive bias is a well-documented automatic process that can have serious negative consequences in a variety of settings. For example, cognitive bias within a forensic science setting can lead to examiners' judgements being swayed by details that they have learned while working on the case, and which go beyond the physical evidence being examined. Although cognitive bias has been studied in many forensic disciplines, such as fingerprints, bullet comparison, and document examination, knowledge of cognitive bias within forensic toxicology is lacking. Here, we address this knowledge gap by assessing the reported use of contextual information by an international group of forensic toxicologists attending the 54th conference of The International Association of Forensic Toxicologists (TIAFT) in Brisbane in 2016. In a first study, participants read a set of simple post-mortem toxicology results (two drug concentrations in blood) and then indicated what information they would normally use when interpreting these results in their day-to-day casework. Using a questionnaire, we then surveyed the familiarity of toxicologists with contextual bias and captured any suggested bias-minimizing procedures for use in forensic toxicology laboratories. Thirty-six participants from 23 different countries and with a range of 1–35 years' forensic toxicology reporting experience volunteered. Analysis of their responses showed that the majority of participants reported using some contextual information in their interpretation of these post-mortem toxicology results (range = 3–15 pieces of information, median ± SD = 11 ± 3), the most common being the deceased's history of prescription or illicit drug use. More than three-quarters of participants reported being familiar with the concept of contextual bias, although few (n = 9) worked in laboratories that had a formal policy covering it. Over half of participants knew of at least one bias-minimizing procedure specifically for forensic toxicology casework, but only a quarter (overall) reported using bias-minimizing procedures in their laboratories. Our results provide substantial evidence that although practising forensic toxicologists are familiar with contextual bias, many report that they still engage in behaviours that could lead to cognitive bias (e.g., through the use of contextual information, through lack of explicit policies or bias-minimizing procedures). We anticipate that our work will form the basis of further research involving a larger sample of participants and examining other potentially relevant factors such as sex/gender, country and accreditation of laboratories.

The Important Role of Contextual Information in Speech Perception in Cochlear Implant Users and Its Consequences in Speech Tests.

This study investigated the role of contextual information in speech intelligibility, the influence of verbal working memory on the use of contextual information, and the suitability of an ecologically valid sentence test containing contextual information, compared with a CNC (Consonant-Nucleus-Consonant) word test, in cochlear implant (CI) users. Speech intelligibility performance was assessed in 50 postlingual adult CI users on sentence lists and on CNC word lists. Results were compared with a normal-hearing (NH) group. The influence of contextual information was calculated from three different context models. Working memory capacity was measured with a Reading Span Test. CI recipients made significantly more use of contextual information in recognition of CNC words and sentences than NH listeners. Their use of contextual information in sentences was related to verbal working memory capacity but not to age, indicating that the ability to use context is dependent on cognitive abilities, regardless of age. The presence of context in sentences enhanced the sensitivity to differences in sensory bottom-up information but also increased the risk of a ceiling effect. A sentence test appeared to be suitable in CI users if word scoring is used and noise is added for the best performers.