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.

Site-directed gene integration in transgenic zebrafish mediated by cre recombinase using a combination of mutant Lox sites

With current gene-transfer techniques in fish, insertion of DNA into the genome occurs randomly and in many instances at multiple sites. Associated position effects, copy number differences, and multiple gene interactions make gene expression experiments difficult to interpret and fish phenotype less predictable. To meet different fish engineering needs, we describe here a gene targeting model in zebrafish. At first, four target zebrafish lines, each harboring a single genomic lox71 target site, were generated by zebrafish transgenesis. The zygotes of transgenic zebrafish lines were coinjected with capped Cre mRNA and a knockin vector pZklox66RFP. Site-specific integration event happened from one target zebrafish line. In this line two integrant zebrafish were obtained from more than 80,000 targeted embryos (integrating efficiency about 10(-4) to 10(-5)) and confirmed to have a sole copy of the integrating DNA at the target genome site. Genomic polymerase chain reaction analysis and DNA sequencing verified the correct gene target events where lox71 and lox66 have accurately recombined into double mutant lox72 and wild-type loxP. Each integrant zebrafish chosen for analysis harbored the transgene rfp at the designated egfp concatenates. Although the Cre-mediated recombination is site specific, it is dependent on a randomly placed target site. That is, a genomic target cannot be preselected for integration based solely on its sequence. Conclusively, an rfp reporter gene was successfully inserted into the egfp target locus of zebrafish genome by Cre-lox-mediated recombination. This site-directed knockin system using the lox71/lox66 combination should be a promising gene-targeting platform serving various purposes in fish genetic engineering.

Is there a preferred classifier for operational thematic mapping?

The importance of properly exploiting a classifier's inherent geometric characteristics when developing a classification methodology is emphasized as a prerequisite to achieving near optimal performance when carrying out thematic mapping. When used properly, it is argued that the long-standing maximum likelihood approach and the more recent support vector machine can perform comparably. Both contain the flexibility to segment the spectral domain in such a manner as to match inherent class separations in the data, as do most reasonable classifiers. The choice of which classifier to use in practice is determined largely by preference and related considerations, such as ease of training, multiclass capabilities, and classification cost. © 1980-2012 IEEE.

Correcting the systematic error of the density functional theory calculation: the alternate combination approach of genetic algorithm and neural network

The alternate combinational approach of genetic algorithm and neural network (AGANN) has been presented to correct the systematic error of the density functional theory (DFT) calculation. It treats the DFT as a black box and models the error through external statistical information. As a demonstration, the AGANN method has been applied in the correction of the lattice energies from the DFT calculation for 72 metal halides and hydrides. Through the AGANN correction, the mean absolute value of the relative errors of the calculated lattice energies to the experimental values decreases from 4.93% to 1.20% in the testing set. For comparison, the neural network approach reduces the mean value to 2.56%. And for the common combinational approach of genetic algorithm and neural network, the value drops to 2.15%. The multiple linear regression method almost has no correction effect here.

Comparison and combination of several stress relief methods for cubic boron nitride films deposited by ion beam assisted deposition

Cubic boron nitride (c-BN) films were prepared by ion beam assisted deposition (IBAD) technique, and the stresses were primary estimated by measuring the frequency shifts in the infrared-absorption peaks of c-BN samples. To test the possible effects of other factors, dependencies of the c-BN transversal optical mode position on film thickness and c-BN content were investigated. Several methods for reducing the stress of c-BN films including annealing, high temperature deposition, two-stage process, and the addition of a small amount of Si were studied, in which the c-BN films with similar thickness and cubic phase content were used to evaluate the effects of the various stress relief methods. It was shown that all the methods can reduce the stress in c-BN films to various extents. Especially, the incorporation of a small amount of Si (2.3 at.%) can result in a remarkable stress relief from 8.4 to similar to 3.6 GPa whereas the c-BN content is nearly unaffected, although a slight degradation of the c-BN crystallinity is observed. The stress can be further reduced down below I GPa by combination of the addition of Si with the two-stage deposition process. (c) 2008 Elsevier B.V. All rights reserved.

Optical characteristics of InAs quantum dots capped with short period GaAs/InAs superlattices and InGaAs combination layers

Long wavelength light emission was realized by capping InAs quantum dots (QDs) with short period GaAs/InAs superlattices (SLs) and an InGaAs strain-reducing layer (SRL). The optical properties were systematically investigated by photoluminescence tests. With increasing the periods of SLs, the emission wavelength of InAs QDs shifts from 1.27 to 1.53 mum. We explain the redshift as a result of the increased QD height with the SLs and the reduced strain in the dot caused by InGaAs SRL. (C) 2004 Published by Elsevier Ltd.

Luminescence properties of self-assembled InAs/GaAs quantum dots covered by InAlAs and InGaAs combination strain-reducing layer

We report the photoluminescence (PL) and structural properties of self-assembled InAs/GaAs quantum dots (QDs) covered by In0.2Al0.8As and In0.2Ga0.8As combination strain-reducing layer (SRL). By introducing a thin InAlAs layer, the ground state emission wavelength redshifts, and the energy splitting between the ground and first-excited states increases to 85 meV at 10 K. The energy splitting further increases to 92 meV and the temperature dependence of full width at half maximum (FWHM) changes for QDs with different SRL after the multi-stacking. These results are attributed to the fact that the combination layer has different effects on QDs compared to the InGaAs SRL.

Effect of In0.2Ga0.8As and In0.2Al0.8As combination layer on band offsets of InAs quantum dots

We demonstrate the self-organized InAs quantum dots capped with thin and In0.2Al0.8As and In0.2Ga0.8As combination layers with a large ground and first excited energy separation emission at 1.35 mum at room temperature. Deep level transient spectroscopy is used to obtain quantitative information on emission activation energies and capture barriers for electrons and holes. For this system, the emission activation energies are larger than those for InAs/GaAs quantum dots. With the properties of wide energy separation and deep emission activation energies, self-organized InAs quantum dots capped with In0.2Al0.8As and In0.2Ga0.8As combination layers are one of the promising epitaxial structures of 1.3 mum quantum dot devices. (C) 2004 American Institute of Physics.

Effect of the InAlAs and InGaAs combination strain-reducing layer on 1.3 mu m emission self-assembled InAs/GaAs quantum dots

Self-assembled InAs quantum dots (QDs) with differing deposition thicknesses covered by InxAl1-xAs (x = 0.2, 0.3) and In0.2Ga0.8As combination strain-reducing layers (CSRLs) were grown by molecular beam epitaxy. Their structural and optical properties were investigated by atomic force microscopy and photoluminescence spectroscopy, respectively. The emission peak position of InAs QDs capped by CSRL can reach 1.34 mum at room temperature with a relatively larger energy splitting of 93 meV between the ground and first excited states.

Photoluminescence of self-assembled InAs/GaAs quantum dots covered by InAlAs and InGaAs combination strain-reducing layer

Self-assembled InAs/GaAs quantum dots covered by the 1-nm InxAl(1-x)As (x = 0.2,0.3) and 3-nm In0.2Ga0.8As combination strain-reducing layer are fabricated, whose height can take up to 30-46 nm. The luminescence emission at a long-wavelength of 1.33 mum and the energy separation between the ground and the first-excited state of 86 meV are observed at room temperature. Furthermore, comparative study proves that the energy separation can increase to 91 meV by multiple stacking.

Abnormal temperature dependence of photoluminescence from self-assembled InAs quantum dots covered by an InAlAs/InGaAs combination layer

In this report we have investigated the temperature dependence of photoluminescence (PL) from self-assembled InAs quantum dots (QDs) covered by an InAlAs/InGaAs combination layer. The ground state experiences an abnormal variation of PL linewidth from 15 K up to room temperature. Meanwhile, the PL integrated intensity ratio of the first excited state to the ground state for InAs QDs unexpectedly decreases with increasing temperature, which we attribute to the phonon bottleneck effect. We believe that these experimental results are closely related to the partially coupled quantum dots system and the large energy separation between the ground and the first excited states. (C) 2003 Elsevier Science Ltd. All rights reserved.

Photoluminescence study of self-assembled InAs/GaAs quantum dots covered by an InAlAs and InGaAs combination layer

We have fabricated a quantum dot (QD) structure for long-wavelength temperature-insensitive semiconductor laser by introducing a combined InAlAs and InGaAs overgrowth layer on InAs/GaAs QDs. We found that QDs formed on GaAs (100) substrate by InAs deposition followed by the InAlAs and InGaAs combination layer demonstrate two effects: one is the photoluminescence peak redshift towards 1.35 mum at room temperature, the other is that the energy separation between the ground and first excited states can be up to 103 meV. These results are attributed to the fact that InAs/GaAs intermixing caused by In segregation at substrate temperature of 520 degreesC can be considerably suppressed by the thin InAlAs layer and the strain in the quantum dots can be reduced by the combined InAlAs and InGaAs layer. (C) 2002 American Institute of Physics.

Tuning of emission wavelength of InAs/GaAs quantum dots sandwiched by combination layers

We investigate about controlling of photoluminescence (PL) wavelengths of InAs/GaAs self-assembled quantum dots (QDs) sandwiched with combination strained-buffer layer (CSBL) and combination strained-reducing layer (CSRL). The emission peak position of QDs is red-shifted to 1.37 mu m. The density of the QDs is increased to 1.17x10(10) cm(-2). It is indicated that optical properties of QDs could be improved by optimizing of the buffer and covering layers for the QDs. These results may provide a new way to further developing GaAs-based 1.3 mu m light sources.