融合依存句法的深度学习藏语句子分割研究

藏语句子分割是藏语自然语言处理中的一项重要且基础性的研究工作。该文根据藏语句子结构特征,在分析藏语句子分割规则与难点的基础上,提出一种融合依存句法的藏语句子分割模型。该模型首先通过词嵌入和藏语依存句法信息嵌入将输入序列映射成实值向量;然后构建融合藏语依存句法的双向LSTM,拼接词语和句法信息特征,提高上下文时序特征的学习能力;最后利用CRF预测出最佳句子分割点。通过对比实验,验证了该模型对藏语句子分割的有效性。实验结果表明,该模型的F₁值为99.4%。     

基于阈值自适应忆阻器Hopfield神经网络的关联规则挖掘算法

针对基于Hopfield神经网络的最大频繁项集挖掘（HNNMFI）算法存在的挖掘结果不准确的问题,提出基于电流阈值自适应忆阻器（TEAM）模型的Hopfield神经网络的改进关联规则挖掘算法。首先,使用TEAM模型设计实现突触,利用阈值忆阻器的忆阻值随方波电压连续变化的能力来设定和更新突触权值,自适应关联规则挖掘算法的输入。其次,改进原算法的能量函数以对齐标准能量函数,并用忆阻值表示权值,放大权值和偏置。最后,设计由最大频繁项集生成关联规则的算法。使用10组大小在30以内的随机事务集进行1000次仿真实验,实验结果表明,与HNNMFI算法相比,所提算法在关联挖掘结果准确率上提高33.9个百分点以上,说明忆阻器能够有效提高Hopfield神经网络在关联规则挖掘中的结果准确率。     

基于规则与统计相结合的藏文文本自动查错方法研究

针对目前藏文文本自动查错方法的不足,该文提出了一种基于规则和统计相结合的自动查错方法.首先以藏文拼写文法为基础,结合形式语言与自动机理论,构造37种确定型有限自动机识别现代藏文字;然后利用查找字典的方法识别梵音藏文字;最后利用互信息和t-测试差等统计方法查找藏语词语搭配错误和语法错误等真字词错误,实现藏文文本的自动查错...     

Design and evaluation of a lyophilized liposomal gel of an antiviral drug for intravaginal delivery

Liposomes of antiviral drug(acyclovir) prepared by rotary evaporation method were incorporated into two bioadhesive polymers, carbopol and HPMC and freeze dried to obtain a unit dosage form. The liposomes, liposomal gels and freeze dried rods were evaluated for various parameters. TEM analysis showed the formation of unilamellar liposomes with a mean diameter ranging from 0.9 μm to 1.2 μm. As the cholesterol content increases from 0.5% to 2%w/w, the entrapment efficiency and vesicle size increased. Carbopol gels exhibited higher viscosity, spreadability, mucoadhesiveness than HPMC gels. The redipsersion of freeze dried forms in SVF was found to be slow and its ex‐vivo retention time was found to be 12 hrs while acyclovir gel retained only for 8.25 hrs. The tablet and gel released 96.93±0.15% acyclovir within 6 hrs and 92.31±0.31% by 8 hrs respectively while freeze dried forms could sustain the release upto 12 hrs. From the stability studies the optimum storage condition was found to be 4‐8°C. © 2013 Wiley Periodicals, Inc. J. Appl. Polym. Sci. 2014 , 131, 39804.     

A catalyst-free growth of aluminum-doped ZnO nanorods by thermal evaporation

The growth of Al:ZnO nanorods on a silicon substrate using a low-temperature thermal evaporation method is reported. The samples were fabricated within a horizontal quartz tube under controlled supply of O₂ gas where Zn and Al powders were previously mixed and heated at 700°C. This allows the reactant vapors to deposit onto the substrate placed vertically above the source materials. Both the undoped and doped samples were characterized using scanning electron microscopy (SEM), field emission scanning electron microscopy (FESEM), energy-dispersive X-ray spectroscopy (EDX), high-resolution transmission electron microscopy (HRTEM) and photoluminescence (PL) measurements. It was observed that randomly oriented nanowires were formed with varying nanostructures as the dopant concentrations were increased from 0.6 at.% to 11.3 at.% with the appearance of ‘pencil-like’ shape at 2.4 at.%, measuring between 260 to 350 nm and 720 nm in diameter and length, respectively. The HRTEM images revealed nanorods fringes of 0.46 nm wide, an equivalent to the lattice constant of ZnO and correspond to the (0001) fringes with regard to the growth direction. The as-prepared Al:ZnO samples exhibited a strong UV emission band located at approximately 389 nm (E _g  = 3.19 eV) with multiple other low intensity peaks appeared at wavelengths greater than 400 nm contributed by oxygen vacancies. The results showed the importance of Al doping that played an important role on the morphology and optical properties of ZnO nanostructures. This may led to potential nanodevices in sensor and biological applications.     

基于基字识别的藏文音节字检错算法研究

在藏文音节字构件中基字占有重要的地位,只要准确的识别出基字的位置,可通过音节字的长度来获得其他构件,并以基字为核心来判断组成音节宇其他构件的搭配是否符合规则。本文通过研究藏文音节字构件组合语法,提出了一种基于基字识别的藏文音节字检错算法。该方法不使用藏文音节字词典及大规模语料库的支撑。通过实验,该方法自动检错藏文音节字的正确率为97％。     

基于CRF的藏文地名识别技术研究

藏文地名识别是藏文命名实体识别中必须要解决的问题。通过分析藏文地名的特点及识别难点,阐述了藏文地名的音节、触发词、地名后续词和格助词等特性适用基于CRF模型的地名识别,通过实验,验证了6种特征对藏文地名识别的有效性。实验结果表明该方法对藏文地名识别的准确率、召回率和[F]值分别达到了96.12%、81.92%和88.45%,实验结果与已有的系统相比,取得了较好的效果。     

The Statistical Evaluation of Categorical Measurements: “Simple Scales,but Treacherous Complexity Underneath”

ABSTRACT

The statistical evaluation of measurements on categorical scales is hampered by hiatuses in insight and conceptualization. Categorical scales have a simple mathematical structure. The underlying empirical reality, however, that they aim to reflect usually has a very complex structure. This complexity induces intricate challenges for the statistical evaluation of the performance of categorical measurement systems. Most current techniques deal ineffectively with these challenges, relying on simplistic conditional independence assumptions and careless sampling strategies. Moreover, they typically evaluate measurement systems in terms of concepts not clearly related to a notion of measurement error. This article proposes an approach for modeling the behavior of categorical measurements based on characteristic curves. The approach is intended to facilitate the development of more effective techniques. It is applied in a case study that illustrates what the authors believe is a realistic degree of complexity.     

Magneto-optical properties of transition metal compounds XPt3 (X = V, Cr, Mn, Fe, Co) and X3Pt (X = Fe, Co)

We have studied theoretically the electronic structure and magneto-optical Kerr effect (MOKE) of XPt₃ (X = V, Cr, Mn, Fe, Co) and X₃Pt (X = Fe, Co) compounds from first principles using the full potential linearized augmented plane wave (FPLAPW) method within the local spin density approximation. The calculated magneto-optical spectra show the features of the experimental spectra which can be attributed to electronic transitions. For the Pt based compounds of 3d transition metals, we find the largest Kerr angle of −1.17° for MnPt₃ in the polar geometry which is in good agreement with the experimental data. Our calculations show an increase in Kerr rotation for XPt₃ series which can be correlated to an increase in the unpaired valence d-electrons of 3d transition metal (X) up to MnPt₃ and then a decrease for FePt₃ and CoPt₃. The Kerr spectrum is analyzed to ascertain the origin of the various peaks in terms of optical transitions. In addition, the effect of interchanging X and Pt atoms in XPt₃ is also studied in order to have greater insight for the dependence of Kerr spectra on configuration and local environment of these two atoms.