基于桥连接的词典学习方法的语义解析

现阶段的语义解析方法大部分都基于组合语义,这类方法的核心就是词典。词典是词汇的集合,词汇定义了自然语言句子中词语到知识库本体中谓词的映射。语义解析一直面临着词典中词汇覆盖度不够的问题。针对此问题,该文在现有工作的基础上,提出了基于桥连接的词典学习方法,该方法能够在训练中自动引入新的词汇并加以学习,为了进一步提高新学习到的词汇的准确度,该文设计了新的词语—二元谓词的特征模板,并使用基于投票机制的核心词典获取方法。该文在两个公开数据集(WebQuestions和Free917)上进行了对比实验,实验结果表明,该文方法能够学习到新的词汇,提高词汇的覆盖度,进而提升语义解析系统的性能,特别是召回率。     

基于DTW改进算法的孤立词识别系统的仿真与分析

传统的DTW算法在进行孤立词语音识别时着重于时间规整和语音测度的计算，而没有对数据的可靠性和有效性进行分析。本文提出了一种改进的端点检测算法，并采用一种改进的DTW算法，在计算机上进行了仿真。实验结果表明采用改进后的DTW算法有效的降低了识别时间和存储数据量，提高了系统性能。     

针对不同地质目标的叠前时间偏移成像解释评价

随着计算机硬件和地震勘探成像技术的发展，叠前时间偏移正逐步替代常规的NMO加DMO加叠后时间偏移成为地震数据成像处理方法的主流。但对于不同的地质目标，叠前时间偏移的成像效果是否优于常规NMO加DMO加叠后时间偏移的成像效果呢？为此，从地震数据成像处理方法、处理流程和处理参数等方面进行了讨论，并基于某地区三维数据常规处理结果和叠前时间偏移处理结果，针对不同地质目标进行了剖析与评价。认为：叠前时间偏移成像的垂向分辨率较常规处理明显降低，但对于空间波阻抗变化明显的河流和断层，叠前时间偏移成像的空间分辨率要高于常规处理；对于小于1／4波长的叠置薄储层，叠前时间偏移成像的垂向和空间分辨率低于常规处理结果。     

2.4GHz动态CMOS分频器的设计

对现阶段的主流高速CMOS分频器进行分析和比较,在此基础上设计一种采用TSPC(truesingle phase clock)和E-TSPC(extended TSPC)技术的前置双模分频器电路.该分频器大大提高了工作频率,采用0.6μm CMOS工艺参数进行仿真的结果表明,在5V电源电压下,最高频率达到3GHz,功耗仅为8mW.     

Monte Carlo simulation of gas phase polymerization of 1,3‐butadiene. Part II: Kinetics optimization and confirmation

Studies on the deactivations and initiations of gas phase polymerizations of 1,3‐butadiene have been achieved by Monte Carlo simulation. Initiation and deactivation control the reaction before and after the peak of the polymerization rate, respectively. The influence of polymerization temperature has been studied. Monte Carlo modeling of polymerization kinetics and mechanism was confirmed by the agreement of experimental data and simulation results of polymerizations run with a temporary evacuation of monomer. The balance of catalysts and active chains is established by both initiation and chain transfer reactions with cocatalyst, which causes a ‘pseudo‐stability’ stage. © 2003 Society of Chemical Industry     

On trading wavelengths with fibers: a cost-performance based study

We consider the effect of multiple fibers on wavelength division multiplexing networks without wavelength conversion. We study networks with dynamic wavelength routing and develop accurate analytical models to compare various possible options using single- and multiple-fiber networks. We use results of an analytical model and simulation-based studies to evaluate the blocking performance and cost of multifiber networks. The number of fibers required providing high performance in multifiber networks and their costs are compared. A case is made for using multiple fibers in each link with fewer wavelengths instead of using a single fiber with many wavelengths. In particular, we show that a network with four fibers per link and with four wavelengths on each fiber without any wavelength conversion on any node yields similar same performance as the networks with one fiber per link and 16 wavelengths per fiber on each link and with full wavelength conversion capability on all nodes. In addition, the multifiber network may also offer the cost advantage depending on the relative cost of components. We develop a parametric cost model to show that multiple fibers in each link are an attractive option. Finally, such multifiber networks also has fault tolerance, with respect to a single fiber failure, already built into the system.     

基于可靠性评价的含缺陷核压力管道疲劳寿命预测方法

含缺陷压力管道经缺陷评定合乎使用后,其疲劳寿命的估计具有重要的工程意义。将整个含缺陷压力管道作为一个整体,分析了含缺陷压力管道的疲劳裂纹扩展特点,提出了相应的含缺陷压力管道疲劳寿命的计算过程,并在基于可靠性评价的基础上,给出了核压力管道的可接受失效概率,最终得到了含缺陷核压力管道疲劳寿命预测方法。     

利用补充分组编码降低MIMO-OFDM系统峰均比

多输入多输出正交频分复用（MIMO—OH）M）系统已成为第四代移动通信系统研究中的热点，而OFDM系统产生的高峰均功率比（PAPR）是限制MIMO—OFDM技术实用化的主要障碍之一。详细介绍了补充分组编码技术的基本原理，并通过仿真与测试的结论与其他编码技术进行了比较，验证了此编码技术对降低峰均功率比确实具有很好的效果。