Metonymy resolution (MR) is a challenging task in the field of natural language processing. The task of MR aims to identify the metonymic usage of a word that employs an entity name to refer to another target entity. Recent BERT-based methods yield state-of-the-art performances. However, they neither make full use of the entity information nor explicitly consider syntactic structure. In contrast, in this paper, we argue that the metonymic process should be completed in a collaborative manner, relying on both lexical semantics and syntactic structure (syntax). This paper proposes a novel approach to enhancing BERT-based MR models with hard and soft syntactic constraints by using different types of convolutional neural networks to model dependency parse trees. Experimental results on benchmark datasets (e.g., ReLocaR, SemEval 2007 and WiMCor) confirm that leveraging syntactic information into fine pre-trained language models benefits MR tasks.

In this study, a high-energy, temporally shaped picosecond ultraviolet (UV) laser running at 100 Hz is demonstrated, with its pulses boosted to 120 mJ by cascaded regenerative and double-pass amplifiers, resulting in a gain of more than 108. With precise manipulation and optimization, the amplified laser pulses were flat-top in the temporal and spatial domains to maintain high filling factors, which significantly improved the conversion efficiency of the subsequent third harmonic generation (THG). Finally, 91 mJ, 470 ps pulses were obtained at 355 nm, corresponding to a conversion efficiency as high as 76%, which, as far as we are aware of, is the highest THG efficiency for a high-repetition-rate picosecond laser. In addition, the energy stability of the UV laser is better than 1.07% (root mean square), which makes this laser an attractive source for a variety of fields including laser conditioning and micro-fabrication.

Stable isotope ratios (δ18O and δD) in Antarctic snow and ice are basic proxy indices of climate in ice core studies. The relation between the ratios has important indicative significance for moisture sources. In general, the fractionation characteristics of the two isotopes vary with different meteorological and topographical conditions. This paper presents the spatial and temporal distribution of meteoric water line (MWL) slopes along a traverse from the Zhongshan Station (ZSS) to Dome A in East Antarctica. It is found that the slopes decrease with the increasing distance inland from the coast and the lowest slope occurred at Dome A, where the long-range transported moisture dominates and clear sky snowing have an influence. The slopes in different layers of the snowpack showed a decreasing trend with depth and this is attributed to the fractionation during the interstitial sublimation and re-condensation processes of the water vapor. Frost flower development on the interior plateau surface can greatly alter the depth evolution of the MWL slope. The coastal snow pits also go through the post-depositional smoothing effect, but their influences are not so significant as the inland regions.

A liquid–gas interface (LGI) on submerged microstructure surfaces has the potential to achieve large slip velocities, which is significant for underwater applications such as drag reduction. However, surfactants adsorbing on the LGI can cause surface tension gradient against the mainstream, which weakens the flow near the LGI and severely limits drag reduction. The mechanism of the effect of surfactants on two-dimensional flows has already been proposed, while the effect of surfactants on the three-dimensional flow near the LGI is still not clear. In our study, we specifically design an experimental system to directly observe a three-dimensional backflow at the LGI. The formation as well as the behaviour of the backflow are demonstrated to be significantly influenced by the surfactant. Combining experimental measurements, theoretical analyses and numerical simulations, we reveal the underlying mechanism of the backflow, which is a competition between the mainstream and the Marangoni flows generated by the interfacial concentration gradients of surfactant simultaneously in streamwise and spanwise directions, reflecting the three-dimensional feature of the backflow. In addition, a kinematic similarity is obtained to characterize the backflow. The current work provides a model system for investigating the three-dimensional backflow at the LGI with surfactants, which is significant for practical applications such as drag reduction and superhydrophobicity.

To reduce the seed length while maintaining the advantages of the cuboid KDP-type crystal, a long-seed KDP crystal with size $471~\text{mm}\times 480~\text{mm}\times 400~\text{mm}$ is rapidly grown. With almost the same high cutting efficiency to obtain third harmonic generation oriented samples, this long-seed KDP-type crystal can be grown with a shorter seed than that of the cuboid KDP-type crystal. The full width at half maximum of the high-resolution X-ray diffraction of the (200) crystalline face is 28.8 arc seconds, indicating that the long-seed KDP crystal has good crystalline quality. In the wavelength range of 377–1022 nm, the transmittance of the long-seed KDP crystal is higher than 90%. The fluence for the 50% probability of laser-induced damage (LID) is $18.5~\text{J}/\text{cm}^{2}$ (3 ns, 355 nm). Several test points survive when the laser fluence exceeds $30~\text{J}/\text{cm}^{2}$ (3 ns, 355 nm), indicating the good LID performance of the long-seed KDP crystal. At present, the growth of a long-seed DKDP crystal is under way.

Self-powered smart systems utilizing the high voltages generated by triboelectric nanogenerators (TENGs) have been systematically reviewed, with several featured applications highlighted, including electrospray, optical device, microplasma, and microfluidic.

To provide a sustainable power solution for electronics, triboelectric nanogenerator (TENG) has been developed since 2012 for high-efficiency mechanical energy harvesting from the ambient environment. TENG has very unique output characteristics including high voltage and limited current density. Thus, it is challenging to directly power most of the commercial electronics in high efficiency. However, these features also bring opportunities for high-voltage applications. Here, we will review the efforts for developing TENG as a controllable high-voltage power source for various applications. The review article will start from fundamental studies about how the high-voltage output was generated, and then, several representative research in recent years will be reviewed, including electrospray, optical devices, microplasma, field emission, and electrically responsive materials. These studies will drive the further development of TENG technology for broad applications and industrializations toward high-efficiency self-powered systems.

A new simple mathematical method has been proposed to predict rock stress around a noncircular tunnel and the method is calibrated and validated with a numerical model. It can be found that the tunnel shapes and polar angles affect the applicable zone of the theoretical model significantly and the applicable zone of a rectangular tunnel was obtained using this method. The method can be used to predict the values of the concentrated stress, and to analyze the change rate of rock stress and back to calculate the mechanical boundary condition in the applicable zone. The results of the stress change rate indicate that the horizontal stress is negatively related to the vertical boundary load and positively related to the horizontal boundary load. The vertical stress is negatively related to the horizontal boundary load and positively related to the vertical boundary load. These findings can be used to explain the evolution of the vertical increment in stress obtained with field-based borehole stress monitoring.

We consider a distributed optimization problem over a multi-agent network, in which the sum of several local convex objective functions is minimized subject to global convex inequality constraints. We first transform the constrained optimization problem to an unconstrained one, using the exact penalty function method. Our transformed problem has a smaller number of variables and a simpler structure than the existing distributed primal–dual subgradient methods for constrained distributed optimization problems. Using the special structure of this problem, we then propose a distributed proximal-gradient algorithm over a time-changing connectivity network, and establish a convergence rate depending on the number of iterations, the network topology and the number of agents. Although the transformed problem is nonsmooth by nature, our method can still achieve a convergence rate, ${\mathcal{O}}(1/k)$, after $k$ iterations, which is faster than the rate, ${\mathcal{O}}(1/\sqrt{k})$, of existing distributed subgradient-based methods. Simulation experiments on a distributed state estimation problem illustrate the excellent performance of our proposed method.

Fast linear transformer driver (FLTD) has some advantages in repetitive operation compared with traditional pulsed power generators. However, different types of gas switches applied in the field of pulsed power technology in recent years cannot reach the requirements of repetitive operation of FLTD. Therefore, the capability of repetitive operation of a multigap gas switch has been investigated in a circuit similar to the basic discharge loop named as brick in this paper. The switch has been triggered more than 2000 times and the distribution of delay time and switch jitter are analyzed and reported. Also, the self-breakdown voltages of the switch during different segments of the triggered breakdown experiment have been tested. The experimental results indicate that the delay time obeys the Gauss distribution and the jitter of 2000 times of discharge is about 2.3 ns.