基于模式识别的文本知识点深度挖掘方法

针对目前大数据知识获取存在的噪声大的问题,提出了文本知识点深度挖掘方法。首先构建了学术论文创造性特征的“问题,方法,结果”三元组本体模型;其次利用模式识别等技术对学术论文文摘进行统计分析、特征提取、机器学习、模式判定分析;最后对学术论文创造性核心知识的三元组进行深度挖掘。实验结果表明,该方法能大大过滤掉学术文献大数据检索的噪声,便于用户快速定位大型学术文献数据库论文的研究问题,采用的新方法和得到的结果能判断学术论文的阅读价值,并为大数据深度知识挖掘和关联发现研究提供基础。该类方法未见有公开的文献报道,属于一种探索性研究和实验。     

Experimental investigation of phase relationship in Ti–Fe-Hf ternary system

Based on diffusion triple and equilibrated alloy methods, phase relations in the Ti–Fe-Hf ternary system were investigated using the experimental data obtained through the combination of optical microscopy (OM), X-ray diffraction (XRD), scanning electron microscopy (SEM), and electron probe microanalysis (EPMA) techniques. Isothermal sections of the Ti–Fe-Hf system at both 1073 K and 1273 K were well constructed. There are three and four three-phase regions in these two sections, respectively. According to the present results, Hf can dissolve into FeTi at approximately 3.0% at both 1073 K and 1273 K. A continuous solid solution of Fe₂(Ti, Hf) forms between the binary intermediate compounds Fe₂Ti and Fe₂Hf (h). Fe₂Hf(c) and FeHf₂ show large solid solubilities. The solubility of Ti in Fe₂Hf(c) changes from 33.9% at 1073 K to 39.0% at 1273 K, while that of FeHf₂ can reach up to approximately 63.0% at 1273 K. No ternary compound exists in the Ti–Fe-Hf ternary system.     

Three-Body Excitations in Fock-Space Coupled-Cluster: Fourth Order Perturbation Correction to Electron Affinity and Its Relation to Bondonic Formalism

In this paper, we present a formulation of highly correlated Fock-space multi-reference coupled-cluster (FSMRCC) methods, including approximate triples on top of the FSMRCC with singles and doubles, which correct the electron affinities by at least at third and up to the fourth order in perturbation. We discuss various partial fourth-order schemes, which are reliable and yet computationally more efficient than the full fourth-order triples scheme. The third-order scheme is called MRCCSD+T^*(3). We present two approximate fourth-order schemes, MRCCSD+T^*−a(4) and MRCCSD+T^*(4). The results that are presented allow one to choose an appropriate fourth-order scheme, which is less expensive and right for the problem. All these schemes are based on the effective Hamiltonian scheme, and provide a direct calculation of the vertical electron affinities. We apply these schemes to a prototype Li₂ molecule, using four different basis sets, as well as BeO and CH⁺. We have calculated the vertical electron affinities of Li₂ at the geometry of the neutral Li₂ molecule. We also present the vertical ionization potentials of the Li₂ anion at the geometry of the anion ground state. We have also shown how to calculate adiabatic electron affinity, though in that case we lose the advantages of direct calculation. BeO has been examined in two basis sets. For CH⁺, four different basis sets have been used. We have presented the partial fourth-order schemes to the EA in all the basis sets. The results are analyzed to illustrate the importance of triples, as well as highlight computationally efficient partial fourth-order schemes. The choice of the basis set on the electron affinity calculation is also emphasized. Comparisons with available experimental and theoretical results are presented. The general fourth-order schemes, which are conceptually equivalent with the Fock-space multi-reference coupled-cluster singles, doubles, and triplets (MRCCSD+T) methods, based on bondonic formalism, are also presented here in a composed way, for quantum electronic affinity.     

三元组深度哈希学习的司法案例相似匹配方法

在数量庞大的司法案例文书中进行相似案例匹配可以有效地提升司法部门的工作效率。但司法案件文本不仅长,而且文本自身还具有一定程度的结构复杂性,因此司法案例文本匹配与传统自然语言处理任务相比,具有较高的难度。为解决上述问题,本文基于三元组深度哈希学习模型提出了一种司法案例相似匹配方法,首先使用预训练的BERT中文模型分组提取文书的特征;再利用文书三元组相似性关系,训练深度神经网络模型,用于生成文书的哈希码表示;最后,基于文书哈希码的海明距离来判断是否为相似案例。实验结果表明,本文采用哈希学习方法极大地降低了文书特征表示的存储开销,提高了相似案例匹配的速度。     

空间飞行器FADS系统算法及奇异值分析

嵌入式大气数据传感(FADS)系统依靠飞行器前端的压力传感器间接得到飞行大气数据。简要介绍了FADS系统的压力模型,针对最小二乘法迭代稳定性差的缺点,给出了求解大气数据迎角和侧滑角的三点法。对三点法求解过程中出现的奇异值问题采用分类分析法进行分析,并给出各组合间优势互补的解决方案。仿真结果表明奇异值分析结果和解决方案正确可行。     

The sequence of a 15 769 bp segment of Pichia anomala identifies the SEC61 and FBP1 genes and five new open reading frames

We have determined the sequence of a 15 769 bp DNA segment of Pichia anomala. The sequence contains seven complete open reading frames (ORFs) longer than 100 amino acids and a putative tRNA gene. Two of the ORFs code for the well-characterized genes SEC61 (which codes for the core subunit of the ER translocation complex) and FBP1 (encoding fructose-1,6-bisphosphatase). A gene coding for a protein similar to S. cerevisiae YDL054c was found between the two genes. These three genes show a different organization (intermingled triples) in three yeast species: Saccharomyces cerevisiae, Candida albicans and P. anomala. Two out of the four remaining ORFs show weak homology with different proteins from other species and the other two show non-significant similarity with previously sequenced genes. The nucleotide sequence has been submitted to the EMBL database under Accession No. AJ306295.     

基于汉语框架网的问句分析及本体知识库推理

针对受限领域问答系统中的问句理解,提出一种将汉语框架网(Chinese FrameNet,简称CFN)应用于问句分析中,并且以旅游交通领域问句为例分析方法的有效性.首先利用CFN语义角色标注和问句句型相结合的方式制定模板,根据CFN语义分析得到的框架元素与本体三元组特点,制定该模板在旅游本体库中答案抽取的规则.该方法在...     

A stabilized finite volume method for the evolutionary Stokes–Darcy system

In this paper, we propose a stabilized fully discrete finite volume method based on two local Gauss integrals for a non-stationary Stokes–Darcy problem. This stabilized method is free of stabilized parameters and uses the lowest equal-order finite element triples $P1\text{–}P1\text{–}P1$ for approximating the velocity, pressure and hydraulic head of the Stokes–Darcy model. Under a modest time step restriction in relation to physical parameters, we give the stability analysis and the error estimates for the stabilized finite volume scheme by means of a relationship between finite volume and finite element approximations with the lower order elements. Finally, a series of numerical experiments are provided to demonstrate the validity of the theoretical results.     

On the use of thresholds in multi-adjoint concept lattices

The size of the concept lattices increases exponentially from the number of objects and attributes. This situation is more complicated in the fuzzy case, in which the considered carriers to evaluate the objects and attributes, and for the relation, are also taken into account. Hence, it is very important to study mechanisms to reduce the size of fuzzy concept lattices maintaining the main information. One of the most important mechanisms to reduce the size of concept lattices is the use of thresholds in the concept-forming operators. This paper studies this mechanism in the general fuzzy framework of multi-adjoint concept lattices, obtaining interesting properties and consequences.