Language model cross adaptation for LVCSR system combination

State-of-the-art large vocabulary continuous speech recognition (LVCSR) systems often combine outputs from multiple sub-systems that may even be developed at different sites. Cross system adaptation, in which model adaptation is performed using the outputs from another sub-system, can be used as an alternative to hypothesis level combination schemes such as ROVER. Normally cross adaptation is only performed on the acoustic models. However, there are many other levels in LVCSR systems' modelling hierarchy where complimentary features may be exploited, for example, the sub-word and the word level, to further improve cross adaptation based system combination. It is thus interesting to also cross adapt language models (LMs) to capture these additional useful features. In this paper cross adaptation is applied to three forms of language models, a multi-level LM that models both syllable and word sequences, a word level neural network LM, and the linear combination of the two. Significant error rate reductions of 4.0-7.1% relative were obtained over ROVER and acoustic model only cross adaptation when combining a range of Chinese LVCSR sub-systems used in the 2010 and 2011 DARPA GALE evaluations. © 2012 Elsevier Ltd. All rights reserved.

Paraphrastic language models

In natural languages multiple word sequences can represent the same underlying meaning. Only modelling the observed surface word sequence can result in poor context coverage, for example, when using n-gram language models (LM). To handle this issue, this paper presents a novel form of language model, the paraphrastic LM. A phrase level transduction model that is statistically learned from standard text data is used to generate paraphrase variants. LM probabilities are then estimated by maximizing their marginal probability. Significant error rate reductions of 0.5%-0.6% absolute were obtained on a state-ofthe-art conversational telephone speech recognition task using a paraphrastic multi-level LM modelling both word and phrase sequences.

Paraphrastic language models and combination with neural network language models

In natural languages multiple word sequences can represent the same underlying meaning. Only modelling the observed surface word sequence can result in poor context coverage, for example, when using n-gram language models (LM). To handle this issue, paraphrastic LMs were proposed in previous research and successfully applied to a US English conversational telephone speech transcription task. In order to exploit the complementary characteristics of paraphrastic LMs and neural network LMs (NNLM), the combination between the two is investigated in this paper. To investigate paraphrastic LMs' generalization ability to other languages, experiments are conducted on a Mandarin Chinese broadcast speech transcription task. Using a paraphrastic multi-level LM modelling both word and phrase sequences, significant error rate reductions of 0.9% absolute (9% relative) and 0.5% absolute (5% relative) were obtained over the baseline n-gram and NNLM systems respectively, after a combination with word and phrase level NNLMs. © 2013 IEEE.

Determination of trace level microcystins in water using solid-phase extraction and high performance liquid chromatography

This paper presents a method for trace level analysis of microcystins in water using solid-phase extraction and high performance liquid chromatography. The optimized condition enabled the determination of common microcystins at levels as low as 0.02 similar to 0.05 mug/L, and the liner range is from 0.1 mug/L to 50 mug/L. The method has been applied to the analysis of field sample from Dianchi lake.

多级多版本数据库管理系统全局串行化

多级调度应该保证事务历史可串行化,满足多级安全特性,不会引入隐通道,并保证高级别事务不会因为无限等待而"饿死".与其他多级数据管理系统调度机制相比,多级多版本时戳调度机制满足上述要求,但该机制存在两个问题,一是事务可能读旧版本,二是要求调度器是可信进程.提出一种多级多版本全局时戳调度机制(MLS_MVGTO),以及依据事务快照生成其全局时戳的基本步骤.给出了预知只读事务信息时的两种改进方法.MLS_MVGTO机制生成的事务历史可串行化,不引入隐通道等,并且该方法避免引入一个全局可信的调度器,并通过对只读事务的深入分析,允许事务读新版本.

一种实用的安全数据库多级关系模型

安全数据库管理系统的安全策略可以为数据库提供强大的安全功能,现在BLP模型是用得比较普遍的策略模型之一。为了适应BLP模型,采用一种实用的多级关系模型作为安全数据库的基础,这是原有关系模型的扩展。该文描述了多级关系模型的原理和内容,特别对多级关系模型的分解与合并做了详细的阐述。并说明了这个模型适合于在多级安全操作系统上实现多级数据库管理系统。

Smoothing LDA Model for Text Categorization

Abstract. Latent Dirichlet Allocation (LDA) is a document level language model. In general, LDA employ the symmetry Dirichlet distribution as prior of the topic-words’ distributions to implement model smoothing. In this paper, we propose a data-driven smoothing strategy in which probability mass is allocated from smoothing-data to latent variables by the intrinsic inference procedure of LDA. In such a way, the arbitrariness of choosing latent variables'priors for the multi-level graphical model is overcome. Following this data-driven strategy,two concrete methods, Laplacian smoothing and Jelinek-Mercer smoothing, are employed to LDA model. Evaluations on different text categorization collections show data-driven smoothing can significantly improve the performance in balanced and unbalanced corpora.

一个支持可信主体特权最小化的多级安全模型

对已有多级安全模型的可信主体支持进行回顾和分析,提出了DLS(离散标记序列)多级安全模型.该模型将可信主体的生命周期分解为一系列非可信状态,对每一个状态赋予一个敏感标记.可信主体的当前敏感标记等于当前非可信状态的敏感标记,非可信状态的切换由预定义的可信请求事件触发.从而可信主体的当前敏感标记可以根据其应用逻辑而动态调整.同时给出了模型保持系统安全性的安全状态和规则.与Bell模型等可信主体敏感标记范围模型相比,该模型在多级安全的策略范围内实现了可信主体的特权最小化.

First principles based predictions of the toughness of a metal/oxide interface

We describe a first-principles-based strategy to predict the macroscopic toughness of a gamma-Ni(Al)/alpha-Al2O3 interface. Density functional theory calculations are used to ascertain energy changes upon displacing the two materials adjacent to the interface, with relaxation conducted over all atoms located within adjoining rows. Traction/displacernent curves are obtained from derivatives of the energy. Calculations are performed in mode I (opening), mode II (shear) and at a phase angle of 45 degrees. The shear calculations are conducted for displacements along < 110 > and < 112 > of the Ni lattice. A generalized interface potential function is used to characterize the results. Initial fitting to both the shear and normal stress results is required to calibrate the unknowns. Thereafter, consistency is established by using the potential to predict other traction quantities. The potential is incorporated as a traction/displacement function within a cohesive zone model and used to predict the steady-state toughness of the interface. For this purpose, the plasticity of the Ni alloy must be known, including the plasticity length scale. Measurements obtained for a gamma-Ni superalloy are used and the toughness predicted over the full range of mode mixity. Additional results for a range of alloys are used to demonstrate the influences of yield strength and length scale.

Estimation of wind erosion rates by using Cs-137 tracing technique: A case study in Tariat-Xilin Gol transect, Mongolian Plateau

Wind erosion is one of the major environmental problems in semi-arid and arid regions. Here we established the Tariat-Xilin Gol transect from northwest to southeast across the Mongolian Plateau, and selected seven sampling sites along the transect. We then estimated the soil wind erosion rates by using the Cs-137 tracing technique and examined their spatial dynamics. Our results showed that the Cs-137 inventories of sampling sites ranged from 265.63 +/- 44.91 to 1279.54 +/- 166.53 Bq.m(-2), and the wind erosion rates varied from 64.58 to 419.63 t.km(-2).a(-1) accordingly. In the Mongolia section of the transect (from Tariat to Sainshand), the wind erosion rate increased gradually with vegetation type and climatic regimes; the wind erosion process was controlled by physical factors such as annual precipitation and vegetation coverage, etc., and the impact of human activities was negligible. While in the China section of the transect (Inner Mongolia), the wind erosion rates of Xilin Hot and Zhengxiangbai Banner were thrice as much as those of Bayannur of Mongolia, although these three sites were all dominated by typical steppe. Besides the physical factors, higher population density and livestock carrying level should be responsible for the higher wind erosion rates in these two regions of Inner Mongolia.

The regional types of China's floating population: Identification methods and spatial patterns

With the rapid increase of the number and influence of floating population in China, it is urgently needed to understand the regional types of China's floating population and their spatial characteristics. After reviewing the current methods for identifying regional types of floating population, this paper puts forward a new composite-index identification method and its modification version which is consisted of two indexes of the net migration rate and gross migration rate. Then, the traditional single-index and the new composite-index identification methods are empirically tested to explore their spatial patterns and characteristics by using China's 2000 census data at county level. The results show: (1) The composite-index identification method is much better than traditional single-index method because it can measure the migration direction and scale of floating simultaneously, and in particular it can identify the unique regional types of floating population with large scale of immigration and emigration. (2) The modified composite-index identification method, by using the share of a region's certain type of floating population to the total in China as weights, can effectively correct the over- or under-estimated errors due to the rather large or small total population of a region. (3) The spatial patterns of different regional types of China's floating population are closely related to the regional differentiation of their natural environment, population density and socio-economic development level. The three active regional types of floating population are mainly located in the eastern part of China with lower elevation, more than 800 mm precipitation, rather higher population densities and economic development levels.

Dependence of performance of organic light-emitting devices on sheet resistance of indium-tin-oxide anodes

The dependence of the performance of organic light-emitting devices(OLEDs) on the sheet resistance of indium-tin-oxide(ITO) anodes was investigated by measuring the steady state current density brightness voltage characteristics and the electroluminescent spectra. The device with a higher sheet resistance anode shows a lower current density, a lower brightness level, and a higher operation voltage. The electroluminescence(EL) efficiencies of the devices with the same structure but different ITO anodes show more complicated differences. Furthermore, the shift of the light-emitting zone toward the anode was found when an anode with a higher sheet resistance was used. These performance differences are discussed and attributed to the reduction of hole injection and the increase in voltage drop over ITO anode with the increase in sheet resistance.

节能有效的无线传感器网络状态信息收集算法

网络状态信息收集协议既要保证信息收集的准确性、实时性,又要保证协议算法的轻量级特性。为解决上述矛盾问题,提出了一种轻量级的、能量有效的、基于无损聚合的层次分簇数据收集机制(QTBDC)。QTBDC首先对网络节点编码并在节点间建立起一个逻辑层次簇结构,然后利用各个子簇状态数据的相似性和编码的连续性,实现了网内无损聚合。该监测机制使得网络状态信息的收集在不丢失数据细节信息的情况下,数据通信量大大减少。经过仿真分析表明,该方法与现有经典数据收集方法相比,实现了节能,延长了网络的生命期。

实现多数据源集成与互操作的扩展视图方法

文中针对多数据源集成与互操作提出了一种扩展视图方法．其基本思想是：以扩展对象模型（ＥＯＭ）为理论基础，通过扩展视图方法实现对多种数据源中的数据（对象）进行多层抽象与封装，以此满足分布环境下不同层次的应用集成需求．文中还讨论了扩展对象模型（ＥＯＭ）、基于扩展视图方法的集成机制，并结合具体的实例给出了扩展视图方法的形式化描述。

冀东油田柳南地区新近系油藏剩余油形成及分布规律研究

This thesis is based on the research project of Study on the Geological Characteristics and Remaining Oil Distribution Law of Neogene Reservoirs in Liunan Area, which is one of the key research projects set by PetroChina Jidong Oilfield Company in 2006. The determination of remaining oil distribution and its saturation changes are the most important research contents for the development and production modification of oilfields in high water-cut phases. Liunan oilfield, located in Tangshan of Hebei Province geographically and in Gaoliu structural belt of Nanpu sag in Bohai Bay Basin structurally, is one of the earliest fields put into production of Jidong oilfield. Focusing on the development problems encountered during the production of the field, this thesis establishes the fine geological reservoir model through the study of reservoir properties such as fine beds correlation, sedimentary facies, micro structures, micro reservoir architecture, flow units and fluid properties. Using routine method of reservoir engineering and technology of reservoir numerical modeling, remaining oil distribution in the target beds of Liunan area is predicted successfully, while the controling factors of remaining oil distribution are illustrated, and the model of remaining oil distribution for fault-block structure reservoirs is established. Using staged-subdivision reservoir correlation and FZI study, the Strata in Liunan Area is subdivided step by step; oil sand body data-list is recompiled; diagram databases are established; plane and section configuration of monolayer sandstone body, and combination pattern of sandstone bodys are summarized. The study of multi-level staged subdivision for sedimentary micro-facies shows that the Lower member of Minghuazhen formation and the whole Guantao formation in Liunan Area belong to meandering river and braided river sedimentary facies respectively, including 8 micro facies such as after point bar, channel bar, channel, natural levee, crevasse splay, abandoned channel, flood plain and flood basin. Fine 3D geological modeling is performed through the application of advanced software and integration of geological, seismic logging and reservoir engineering data. High resolution numerical simulation is performed with a reserve fitting error less than 3%, an average pressure fitting fluctuation range lower than 2Mpa and an accumulate water cut fitting error less than 5%. In this way, the distribution law of the target reservoir in the study area is basically recognized. Eight major remaining oil distribution models are established after analysis of production status and production features in different blocks and different layers. In addition, fuzzy mathematics method is used to the integreted evaluation and prediction of abundant remaining oil accumulation area in major production beds and key sedimentary time units of the shallow strata in Liunan Area and corresponding modification comments are put forward. In summary, the establishment of fine reservoir geological model, reservoir numerical simulation and distribution prediction of remaining oil make a sound foundation for further stimulation of oilfield development performance.