6664 research outputs found
Sort by
Convolution-deconvolution word embedding: an end-to-end multi-prototype fusion embedding method for natural language processing
Get PDFExisting unsupervised word embedding methods have been proved to be effective to capture latent semantic information on various tasks of Natural Language Processing (NLP). However, existing word representation methods are incapable of tackling both the polysemousunaware and task-unaware problems that are common phenomena in NLP tasks. In this work, we present a novel Convolution-Deconvolution Word Embedding (CDWE), an end-to-end multi-prototype fusion embedding that fuses context-specific information and taskspecific information. To the best of our knowledge, we are the first to extend deconvolution (e.g. convolution transpose), which has been widely used in computer vision, to word embedding generation. We empirically demonstrate the efficiency and generalization ability of CDWE by applying it to two representative tasks in NLP: text classification and machine translation. The models of CDWE significantly outperform the baselines and achieve state-of-the-art results on both tasks. To validate the efficiency of CDWE further, we demonstrate how CDWE solves the polysemous-unaware and task-unaware problems via analyzing the Text Deconvolution Saliency, which is an existing strategy for evaluating the outputs of deconvolution
Prevalence of overweight and obesity among adolescents in Bangladesh: do eating habits and physical activity have a gender differential effect?
Get PDFThe aim of the study is to examine the gender differential outcomes of food habits and physical activities on obesity among school-aged adolescents in Bangladesh. In this study, a nationally representative data extracted from the Global School-based Student Health Survey (GSHS) 2014 was utilized. The information related to physical and mental health was collected from 2989 school-aged adolescents in Bangladesh. To fulfill the aim, an exploratory data analysis and multivariate logistic regression model were employed in this study. The female adolescents’ were at lower risk of overweight/obesity (AOR = 0.573) with a prevalence of 7.4% than males (9.9%) in Bangladesh. The results showed that high consumption of vegetables (both: AOR = 0.454; males: AOR = 0.504; and females: AOR = 0.432), high soft drink consumption (both: AOR=2.357; males: AOR = 2.929; and females: AOR = 1.677), high fast food eating (both: AOR = 2.777; males: AOR = 6.064; and females: AOR = 1.695), sleep disturbance (both: AOR = 0.675; males: AOR = 0.590; and females: AOR = 0.555), and regular walk or bike to school (both: AOR = 0.472; males: AOR = 0.430; and females: AOR = 0.557) was the vital influencing factors of overweight/obesity among adolescents of both sexes in Bangladesh. Besides, fruit eating, vigorous physical activity, physical education attendance, sitting activities were also identified as significant predictors of overweight and obesity for males in Bangladesh. The high practice of fruit and vegetable consumption regularly, avoiding soft drinks and fast food, increasing the vigorous intensity of physical activity, regular attendance in physical education and abating higher sitting activity in leisure time can reduce the risk of overweight/obesity among adolescents of both sexes in Bangladesh
Decentralized document version control using ethereum blockchain and IPFS
Get PDFIn this paper, we propose a blockchain-based solution and framework for document sharing and version control to facilitate multi-user collaboration and track changes in a trusted, secure, and decentralized manner, with no involvement of a centralized trusted entity or third party. This solution is based on utilizing Ethereum smart contracts to govern and regulate the document version control functions among the creators and developers of the document and its validators. Moreover, our solution leverages the benefits of IPFS (InterPlanetary File System) to store documents on a decentralized file system. The proposed solution automates necessary interactions among multiple actors comprising developers and approvers. Smart contracts have been developed using Solidity language, and their functionalities were tested using the Remix IDE (Integrated Development Environment). The paper demonstrates that our smart contract code is free of commonly known security vulnerabilities and attacks. The code has been made publically available at Github
Dr Monique Charles - the after party
No full textBlaze and Big E interview Dr Monique Charles about her interest in music and its impac
"Why does music move us: music, health and wellbeing"
No full textMusic is increasingly seen to have an essential non-pharmacological therapeutic role within healthcare and in particular for the treatment of Post-Traumatic Stress Disorder. As Thaut (2005: 115) has argued, ‘brain research involving music has shown that music has a distinct influence on the brain by stimulating physiologically complex cognitive, affective, and sensorimotor processes. [...] This is a very critical step in the historical understanding of music in therapy and medicine’. In this keynote lecture, Emilie Capulet will present and discuss a wide array of musical activities, including Guided Imagery and Music (Dukic, H., Parncutt, R., & Bunt, L. (2019)) and the neurochemistry of music (Chanda ML, Levitin DJ. ‘The Neurochemistry of Music’. Trends Cogn Sci. 2013. Apr. 17(4):179-93), as well as music therapy activities (Landis-Shack, Nora et al. “Music Therapy for Posttraumatic Stress in Adults: A Theoretical Review.” Psychomusicology vol. 27,4 (2017): 334-342) which have a positive impact on cognitive, emotional, behavioral and physiological levels
Orientation independent chipless RFID tag using novel trefoil resonators
Get PDFIn this paper, a compact and fully passive bit encoding circuit, capable of operating as a chipless radio frequency identification (RFID) tag is presented. The structure consists of novel concentric trefoil-shaped slot resonators realized using Rogers RT/duroid 5880 laminate, occupying a physical footprint of 13.55 13.55 mm^2. Each resonating element is associated with a particular data bit, having a 1:1 resonator-to-bit correspondence. Bit sequences are configured through introducing modifications in the geometric structure either by addition or exclusion of each nested slot resonator. Such changes manifest directly in the electromagnetic signature of the tag as presence or absence of corresponding resonant peaks. The proposed 10-bit tag offers minimized inter-resonator mutual coupling and insensitivity to changes in polarization and incident angles thereby demonstrating orientation independent functionality. Moreover, error-free encoding is achieved through stabilizing the shift in resonant frequencies for a variety of different geometric configurations and orientation of the structure. The tag operates within the license-free ultra-wideband ranging from 5.4 to 10.4 GHz, providing spectral bit capacity and bit density of 2 bits/GHz and 5.44 bits/cm^2 respectively
Inscapes of musical listening: sensing/feeling/imagining
Get PDFThis article explores the ramifications of listening through somatics (the Feldenkrais Method), psychoanalysis, and musical improvisation
Life in between phase: understanding and manipulating microphone relationships with visual tools
Get PDFPhase relationships are a complex and mystifying phenomenon for early stage recording engineers and university students. In this paper, we take the analytical capabilities of time shifting plugin Auto-Align and use it to develop new methods of understanding phase interaction. We utilise the visualisations and time shifting features to assist in recording a multi-miked drum kit and in the post-production soundstage for a thirty-two piece big band. We explore new methodologies for phase interaction and microphone manipulation by running these two experiments and documenting the process using a combination of text, audio, and video
Spotlight on mental health and education: SIG hosts inaugural conference
Get PDFBlog post: Over 80 delegates attended the first conference of the recently launched Mental Health, Wellbeing and Education special interest group (SIG) at the University of West London (UWL) in July. The theme, ‘Researching Mental Health and Education: Where are we now?’, attracted academics, practitioners, students and others working across a range of education settings from early years to adult community learning. The impetus for the event was to provide a much-needed forum to showcase the latest research in the field, facilitate knowledge exchange and share best practices. There were many highlights from a diverse programme which comprised research presentations, roundtable discussions and a panel debate