实体关系抽取的技术方法综述*

对实体关系抽取研究以MUC和ACE评测为主线的发展进行总结,并指出实体关系抽取任务普遍存在的三个问题是特定领域标引数据集的获取、模式的获取以及共指消解。在对当前关系抽取的相关文献、系统和项目进行分析研究的基础上,将基于非结构化文本的实体关系抽取技术方法归纳为：基于模式匹配的关系抽取、基于词典驱动的关系抽取、基于机器学习的关系抽取、基于Ontology的关系抽取以及混合抽取方法,旨在为进一步构建实体关系抽取系统提供良好借鉴。     

车牌字符识别研究

本文对以切割好的并经过归一化处理的车牌字符用K-L变换提取特征向量,K-L变换能降低特征维数并保持字符图象的主要特征。采用了四个径向基神经网络对车牌字符识别,降低了识别复杂度,具有较好的识别效果。     

High-extensible scene graph framework based on component techniques

INTRODUCTION With the tremendous growth of the digital en- tertainment industry in the past decades, the scale of game scene is becoming more gigantic. To effectively manage and organize the massive data, game engine usually uses hierarchical scene graph. Many popular 3D modelling tools, such as Maya, 3D Studio Max, use scene graph to manage the scene content (Sowizral, 2000). 3D modelling software—Houdini directly edits scene graph to create and modify the virtual environment. Most v…     

Role of recognition of prior learning for emerging economies: learning from a four sector pilot project in India

Recognition of prior learning (RPL) has increasingly been recognised as a useful way to certify experience and competencies gained in the workplace informally. Particularly pertinent is RPL in the context of international migration or in economies, where system reforms seek to overcome challenges related to access to formal training and qualifications. India has the goal to skill 300 million of its growing workforce and provide them access at various levels to certification and further learning opportunities. As a part of the overall system reforms, the Government of India, jointly with the ILO, pilot tested RPL in four sectors in collaboration with the industry and built the foundation for wider replication for another 26 growth sectors by undertaking a tracer study and evaluation. This paper can empirically establish that RPL has had a positive effect on income opportunities, occupational safety, social status and openness to further learning. The paper further discusses the challenges encountered during the implementation of the RPL projects before concluding with recommendations, highlighting the aspects required for quality RPL in an emerging economy. It closes by emphasising the importance of conducting tracer studies to determine impact and the need to provide complementary training during the RPL process.     

Association between mothers’ problematic Internet use and maternal recognition of child abuse

BackgroundThere are few studies about mothers’ problematic Internet use (PIU). Mothers’ PIU may lead to inadequate parenting and child abuse.ObjectiveThis cross-sectional study aimed to clarify the association between mothers’ PIU and their recognition of child abuse.Participants and settingWe analyzed data collected of health examinations of children aged 4 months, 1.5 years, and 3 years which were carried out in Matsue City, Shimane Prefecture, Japan between April 2016 and March 2017. The number of the subjects were 1685, 1729, 1674, respectively.MethodsWe used logistic regression analysis to clarify the association between mothers’ PIU (Young’s Diagnostic Questionnaire for Internet Addiction score: ≥5) and their recognition of child abuse (selecting < True of me > for < I sometimes think that I am abusing my child > on a questionnaire survey), which was adjusted for covariates such as maternal age, number of children, daytime caretaker, social support, postpartum depression, and current smoking status of the parents.ResultsBased on the multivariate logistic regression analysis, the mothers’ PIU was significantly correlated with their recognition of child abuse for children aged 4 months, 1.5 years, or 3 years [odds ratio (OR): 13.30, 95% confidence interval (CI): 1.26–139.98, OR: 7.02, 95% CI: 1.28–38.55, and OR: 28.06, 2.48–317.93, respectively].ConclusionThis study revealed the possibility that mothers with PIU recognize child abuse more than mothers without PIU. However, further studies should be conducted to increase reliability and validity.     

学术系统中的社会资本及其效应

学术系统中的社会资本是学者在学术活动中获取和使用的嵌入在学术交流与合作网络中的资源,兼具社会性和学术性,主要表现为毕业院校、师徒关系和任职院校等形式。学术性社会资本既发挥着重要的正面作用,也显示出令人担忧的负面效应。如何面对学术性社会资本以及如何规避它的负面效应就成为优化学术系统运行机制、健全学术制度时不得不予以反思的两个问题。     

“工程识图与辅助设计”混合式教学改革与实践

浙江财经大学“工程识图与辅助设计”为浙江省级线上线下混合式一流本科课程。分析“工程识图与辅助设计”课程的性质、特点以及在传统教学模式下课程教学存在的问题;对照“金课”的“两性一度”标准以及课程教学存在的问题,认为开展线上线下混合式教学改革是“工程识图与辅助设计”课程的建设路径;构建“三体系+三模式+三阶段+三形式”的教学组织形式,包括重组课程内容、运用多种驱动模式、重构课前课中课后教学流程以及形成线上线下混合式有机融合的教学形式,并从网络平台、课程资源与考核方式等方面开展课程建设。实践效果表明,采用线上线下混合式教学有效解决了传统课堂教学存在的问题,并能激发学生的学习兴趣,提升学生学习效果。     

RDA的实体及关系解析

FRBR的实体-关系模型是RDA编制的结构基础,这种概念完全不同于AACR2的理念,同时它也是理解RDA的难点论文主要针对RDA中的实体、实体中的元素、实体之间的关系进行了分析,并利用图表进行详细的说明.结论是：RDA对文献描述的四个层次包括作品、内容表达、具体表现、单件,可以对书目信息进行深度揭示,四个层次之间也存在比较复杂的关系.     

虹膜识别技术发展概述

本文在回顾虹膜识别的研究背景基础上,对国内外虹膜识别方法的研究进展进行了综述,总结了存在的研究难点并提出了今后的发展方向。     

LD夜视成像探测中基于PSF的超分辨率重建技术研究

夜视目标识别在视频监控中起着非常重要的作用。从早期的红外热成像发展至今,激光夜视成像系统被广泛的应用。它不仅受到大气吸收和散射特性的影响,还与光学系统中的透镜及传感器的衍射极限有关。在既有成本下,通过图像复原和超分辨率重建,可以尽可能的提高夜视成像的成像质量。很多学者的研究结果表明,成像系统的点扩散函数作为先验知识,可以提高图像复原的质量。因此,本文在最大后验概率框架下,实现点扩散函数与凸集投影超分辨率重建相结合,并应用到半导体激光夜视成像探测系统中,客观的图像质量评价标准被用来决定迭代次数。目标识别的实验结果表明,用本文提出的重建方法可以有效提高半导体激光夜视成像的范围和成像质量。