Semantic word shifts in a scientific domain

Understanding semantic word shifts in scientific domains is essential for facilitating interdisciplinary communication. Using a data set of published papers in the field of information retrieval (IR), this paper studies the semantic shifts of words in IR based on mining per-word topic distribution over time. We propose that semantic word shifts not only occur over time, but also over topics. The shifts are examined from two perspectives, the topic-level and the context-level. According to the over-time word-topic distribution, stable words and unstable words are recognized. The diverging and converging trends in the unstable type reveal characteristics of the topic evolution process. The context-level shifts are further detected by similarities between word vectors. Our work associates semantic word shifts with the evolving of topics, which facilitates a better understanding of semantic word shifts from both topics and contexts.     

Do funded papers attract more usage?

Research funding has been seen as one of the most important resource in the reward system of science. And usage of publications denotes an interesting perspective of user behavior in scientific communication. This study aims to address the relationship between funding and Usage Count, which is a new metrics item established on the platform of Web of Science. Full records of 300,010 articles published in 2013 were downloaded in October 2015, and divided into six disciplines, including information science library science, education educational research, economics, computer science, materials science, and chemistry. Seven indicators were proposed to measure the impact, including Funding rate, Citation per paper, Usage rate, Usage per paper, Citation difference, Usage difference, and Conversion rate. It concluded funding has impact on usage and citation, and funded papers attract more usage, but varying in different disciplines. Usage Count can be used in the extension of citation metrics but with limits. This study originally engages with usage metrics and detected that there is positive correlation between usage and funding.     

A multi-criteria context-sensitive approach for social image collection summarization

Recent increase in the number of digital photos in the content sharing and social networking websites has created an endless demand for techniques to analyze, navigate, and summarize these images. In this paper, we focus on image collection summarization. Earlier methods in image collection summarization consider representativeness and diversity criteria while recent ones also consider other criteria such as image quality, aesthetic or appeal. In this paper, we propose a multi-criteria context-sensitive approach for social image collection summarization. In the proposed method, two different sets of features are combined while each one looks at different criteria for image collection summarization: social attractiveness features and semantic features. The first feature set considers different aspects that make an image appealing such as image quality, aesthetic, and emotion to create attractiveness score for input images while the second one covers semantic content of images and assigns semantic score to them. We use social network infrastructure to identify attractiveness features and domain ontology for extracting ontology features. The final summarization is provided by integrating the attractiveness and semantic features of input images. The experimental results on a collection of human generated summaries on a set of Flickr images demonstrate the effectiveness of the proposed image collection summarization approach.     

Efficient PL Emission from p-type Porous Silicon: A Comparative Study for Selection of Effective Anodization Parameters

The physical mechanism of highly efficient photoluminescence (PL) emission from p-type silicon is described by a comparative study of the effectiveness of the etching parameters in an electrochemical anodization technique. Two series of porous silicon samples were prepared in a combination of anodization current and time, to maintain the total amount of anodic charge transfer constant. Photoluminescence studies show that irrespective of the amount of charge transfer, the samples prepared with comparatively higher current density show an efficient PL as well as stronger blueshift in the emission energy vis-à-vis the samples prepared for longer durations. An overall decrease in crystallite size, as estimated by Raman spectral analysis, was observed for both series of samples with the progress of charge transfer. Comparative analysis shows a marginal difference in crystallite size for both series of samples in the initial state of charge transfer, whereas major differences arise at higher values. This is explained with the formation of silicon suboxide on the porous surface at higher current density, leading to initiation of side wall reaction, and higher reduction rate in crystallite size as well as strong luminescence due to the carrier quantum confinement effect.     

Image and Feature Space Based Domain Adaptation for Vehicle Detection

The application of deep learning in the field of object detection has experienced much progress. However, due to the domain shift problem, applying an off-the-shelf detector to another domain leads to a significant performance drop. A large number of ground truth labels are required when using another domain to train models, demanding a large amount of human and financial resources. In order to avoid excessive resource requirements and performance drop caused by domain shift, this paper proposes a new domain adaptive approach to cross-domain vehicle detection. Our approach improves the cross-domain vehicle detection model from image space and feature space. We employ objectives of the generative adversarial network and cycle consistency loss for image style transfer in image space. For feature space, we align feature distributions between the source domain and the target domain to improve the detection accuracy. Experiments are carried out using the method with two different datasets, proving that this technique effectively improves the accuracy of vehicle detection in the target domain.