面向小语种机器翻译的平行语料库构建方法

神经机器翻译模型的训练效果在很大程度上取决于平行语料库的规模和质量。除了一些常见语言外,汉语与小语种间高质量平行语料库的建设一直处于滞后状态。现有小语种平行语料库多采用自动句子对齐技术利用网络资源构建而成,在文本质量和领域等方面有诸多局限性。采用人工翻译的方式可以构建高质量平行语料库,但是缺乏相关经验和方法。文中从机器翻译实践者和研究者角度出发,介绍了经济高效的人工构建小语种平行语料库的工作,包括其总体目标、实施过程、流程细节和最后结果。在构建过程中尝试并积累了各种经验,形成了小语种到汉语平行语料库构建方法、建议的总结。最终,成功构建了波斯语到汉语、印地语到汉语、印度尼西亚语到汉语各50万条高质量平行语料。实验结果表明,所构建的平行语料库有较好的质量,提高了小语种神经机器翻译模型的训练效果。     

Hydrogen wettability in carbonate reservoirs: Implication for underground hydrogen storage from geochemical perspective

Hydrogen has been considered as a promising renewable source to replace fossil fuels to meet energy demand and achieve net-zero carbon emission target. Underground hydrogen storage attracts more interest as it shows potential to store hydrogen at large-scale safely and economically. Meanwhile, wettability is one of the most important formation parameters which can affect hydrogen injection rate, reproduction efficiency and storage capacity. However, current knowledge is still very limited on how fluid-rock interactions affect formation wettability at in-situ conditions. In this study, we thus performed geochemical modelling to interpret our previous brine contact angle measurements of H₂-brine-calcite system. The calcite surface potential at various temperatures, pressures and salinities was calculated to predict disjoining pressure. Moreover, the surface species concentrations of calcite and organic stearic acid were estimated to characterize calcite-organic acid electrostatic attractions and thus hydrogen wettability. The results of the study showed that increasing temperature increases the disjoining pressure on calcite surface, which intensifies the repulsion force of H₂ against calcite and increases the hydrophilicity. Increasing salinity decreases the disjoining pressure, leading to more H₂-wet and contact angle increment. Besides, increasing stearic acid concentration remarkably strengthens the adhesion force between calcite and organic acid, which leads to more hydrophobic and H₂-wet. In general, the results from geochemical modelling are consistent with experimental observations that decreasing temperature and increasing salinity and organic acid concentration increase water contact angle. This work also demonstrates the importance of involving geochemical modelling on H₂ wettability assessment during underground hydrogen storage.     

Identification of the Primary Factors Determining the Specificity of Human VKORC1 Recognition by Thioredoxin-Fold Proteins

Redox (reduction–oxidation) reactions control many important biological processes in all organisms, both prokaryotes and eukaryotes. This reaction is usually accomplished by canonical disulphide-based pathways involving a donor enzyme that reduces the oxidised cysteine residues of a target protein, resulting in the cleavage of its disulphide bonds. Focusing on human vitamin K epoxide reductase (hVKORC1) as a target and on four redoxins (protein disulphide isomerase (PDI), endoplasmic reticulum oxidoreductase (ERp18), thioredoxin-related transmembrane protein 1 (Tmx1) and thioredoxin-related transmembrane protein 4 (Tmx4)) as the most probable reducers of VKORC1, a comparative in-silico analysis that concentrates on the similarity and divergence of redoxins in their sequence, secondary and tertiary structure, dynamics, intraprotein interactions and composition of the surface exposed to the target is provided. Similarly, hVKORC1 is analysed in its native state, where two pairs of cysteine residues are covalently linked, forming two disulphide bridges, as a target for Trx-fold proteins. Such analysis is used to derive the putative recognition/binding sites on each isolated protein, and PDI is suggested as the most probable hVKORC1 partner. By probing the alternative orientation of PDI with respect to hVKORC1, the functionally related noncovalent complex formed by hVKORC1 and PDI was found, which is proposed to be a first precursor to probe thiol–disulphide exchange reactions between PDI and hVKORC1.     

Korean automatic spacing using pretrained transformer encoder and analysis

Automatic spacing in Korean is used to correct spacing units in a given input sentence. The demand for automatic spacing has been increasing owing to frequent incorrect spacing in recent media, such as the Internet and mobile networks. Therefore, herein, we propose a transformer encoder that reads a sentence bidirectionally and can be pretrained using an out-of-task corpus. Notably, our model exhibited the highest character accuracy (98.42%) among the existing automatic spacing models for Korean. We experimentally validated the effectiveness of bidirectional encoding and pretraining for automatic spacing in Korean. Moreover, we conclude that pretraining is more important than fine-tuning and data size.     

Failure and fragmentation of ceramic target with varying geometric configuration under ballistic impact

The failure and fragmentation of monolithic bare alumina 99.5% ceramic target and energy dissipation of steel 4340 projectile have been studied in a series of ballistic experiments carried out, with the incidence velocities in a range, 122–290 m/s. The velocity drop and energy dissipation increased with incidence velocity for 10 mm thick target with damage zone extended upon the whole area of rear face at higher velocities. The ballistic results obtained with the 10 mm thick target have been compared with the ballistic performance of the 5 mm thick target used in a previous study to explore the effects of target thickness on the failure mechanism. A model for the residual velocity of projectile after perforation of the single layered ceramic target has been developed based on the Lambert Jonas model by using the experimental data available for 5 mm and 10 mm thick alumina 99.5% target against 10.9 mm projectile. The residual velocities and damage patterns were reproduced with a reasonable amount of accuracy by a three-dimensional finite element model developed on commercial ABAQUS/CAE. The effect of obliquity and projectile diameter to target thickness ratio (D/T) on ballistic performance has been determined by the numerical simulation model with impact velocity in a range of 300–500 m/s. A spatial variation of ejected fragments velocity at different time steps was plotted to develop a velocity profile for the ceramic fragments coming out of the target. A semi-empirical model has been proposed for residual velocity after perforation of a monolithic ceramic target, relating to the incidence velocity and projectile diameter to target thickness ratio. The monolithic ceramic targets have been investigated for a comparative assessment of energy dissipation by the ceramic layer to eventually design an efficient front layer of a ceramic based composite armour in future studies.     

结合预训练模型和语言知识库的文本匹配方法

针对文本匹配任务,该文提出一种大规模预训练模型融合外部语言知识库的方法。该方法在大规模预训练模型的基础上,通过生成基于WordNet的同义—反义词汇知识学习任务和词组—搭配知识学习任务引入外部语言学知识。进而,与MT-DNN多任务学习模型进行联合训练,以进一步提高模型性能。最后利用文本匹配标注数据进行微调。在MRPC和QQP两个公开数据集的实验结果显示,该方法可以在大规模预训练模型和微调的框架基础上,通过引入外部语言知识进行联合训练有效提升文本匹配性能。     

A values–beliefs–attitude model of local food consumption: An empirical study in China and Denmark

Consumers’ and policy makers’ interest in local foods is growing. Accordingly, researchers are also increasingly paying attention to the consumption of local foods. Studies have identified preference for local foods as an emergent consumer ideology called “locavorism”, but they have not yet addressed its antecedents or put it into a theoretical context. In addition, extant research provides several insights into local food buying behaviour in developed economies (e.g., USA, UK, Germany, or Italy); however, studies simultaneously conducted in developed and emerging economies are lacking. To address these research gaps, this study develops a conceptual framework with proposed relationships among values, beliefs (locavorism and fresh start mindset), and attitudes towards and intentions to purchase local foods in China and Denmark. We conducted an online survey in China and Denmark that evaluated our constructs with pre-developed multiple-item measures. Using structural equation modelling to test the integrated model, we find that values and long-term orientation are antecedents of consumer beliefs but the influences of values on consumer beliefs differ between collectivistic-dominated China and individualistic-dominated Denmark. Specifically, collectivistic values are significantly and positively related to locavorism in both countries, while individualistic values are strongly linked to locavorism only in Denmark; collectivistic values have no effects on fresh start mindset for the two samples, but individualistic values are significantly and positively related to a fresh start mindset in Denmark. In addition, consumer beliefs are significantly and positively associated with attitudes towards and intentions to buy local foods. Local food marketers can use our findings to target their communications more effectively.     

2D BASIN MODELLING IN THE EASTERN VARISCAN FOLD BELT (CZECH REPUBLIC): INFLUENCE OF THRUSTING ON PATTERNS OF THERMAL MATURATION

This study evaluates the geothermal history of Palaeozoic sedimentary rocks from the easternmost part of the Variscan external zone in the NE Czech Republic. The objective was to investigate the geothermal history of pre‐Variscan Palaeozoic carbonates in the study area, and to assess its relationship with that of the overlying Variscan flysch. In the study area, the Palaeozoic succession occurs at the surface or is overlain by Miocene sediments of the Carpathian Foredeep. Palaeozoic nappes and the main Variscan overthrust have been documented in the subsurface at the deep Pot?tát‐1borehole. Vitrinite reflectance measurements on 38 samples from the Pot?tát‐1 well and 19 samples from nearby surface outcrops and shallow boreholes were available. A 2D thermal model was created using PetroMod and the thermal maturity evolution was modelled by EASY%Ro. The thermal model was constructed based on interpretations of two NW‐SE seismic profiles (lines 5/83 and 5/84) oriented perpendicular to the main Variscan thrusts. The results were calibrated using measured vitrinite reflectance and were adjusted with 1D models from three shallow boreholes. At the Pot?tát‐1 borehole, modelled maximum palaeo‐temperatures of the Variscan flysch (Moravice Formation) ranged from 310°C at a depth of 7.3 km (the top of the preserved succession) to 395°C at the base of the succession, resulting in thermal maturities of >4%R_r. Peak maturation occurred prior to the end of Variscan thrusting. Modelling suggests that the basal heat flow for these thrust units reached a maximum value of 63 mW/m² at 325 Ma. In addition, the modelling suggests that the maturity of the Palaeozoic carbonates was controlled by the thickness of the overlying Variscan flysch nappes. Maximum palaeo‐temperatures for the Palaeozoic carbonates ranged from 265°C at the top of the interval (at a depth of 7.1 km) to 290°C at the base, resulting in a maturity of 3.8 to >4%R_r which is within the dry gas window. The study suggests that basal heat flows in the original (pre‐thrust) Early Carboniferous sedimentary basin were slightly higher than those in the post‐thrust location for the Variscan flysch nappes. This should be taken into account when evaluating the petroleum system in the South Moravian oil province (SW Czech Republic) where a complete sedimentary sequence has not been preserved.     

Energy efficiency of fibre reinforced soil formation at small element scale: Laboratory and numerical investigation

This paper explores the aspects related to the energy consumption for the compaction of unreinforced and fibre reinforced samples fabricated in the laboratory. It is well known that, for a fixed soil density, the addition of fibres invariably results in an increased resistance to compaction. However, similar peak strength properties of a dense unreinforced sample can be obtained using looser granular soil matrices mixed with small quantities of fibres. Based on both experimental and discrete element modelling (DEM) procedures, this paper demonstrates that less compaction energy is required for building loose fibre reinforced sand samples than for denser unreinforced sand samples while both samples show similar peak strength properties. Beyond corroborating the macro-scale experimental observations, the result of the DEM analyses provides an insight into the local micro-scale mechanisms governing the fibre-grain interaction. These assessments focus on the evolution of the void ratio distribution, re-arrangement of soil particles, mobilisation of stresses in the fibres, and the evolution of the fibre orientation distribution during the stages of compaction.