Search

The encoder-decoder model has achieved remarkable results in natural language generation. However, in the dialogue generation work, we often ignore the influence of the dialogue context information and topic information in the generation, resulting in the generated replies not close to the context or lack of topic information leads to general responses. In this work, we study the generation of multi-turn dialogues based on a large corpus and take advantage of the context information and topic information of the conversation in the process of dialogue generation to generate more coherent context-sensitive responses. We improve upon existing models and attention mechanisms and propose a new hierarchical model to better solve the problem of dialogue context (the HAT model). This method enables the model to obtain more contextual information when processing and improves the ability of the model in terms of contextual relevance to produce high-quality responses. In addition, to address the absence of topics in the responses, we pre-train the LDA(Latent Dirichlet Allocation) topic model to extract topic words of the dialogue content and retain as much topic information of dialogue as possible. Our model is extensively tested in several corpora, and the experiments illustrate that our model is superior to most hierarchical and non-hierarchical models with respect to multiple evaluation metrics.

For the 3D quasilinear wave equation $-\big(1+(\partial_tϕ)^p\big)\partial_t^2ϕ+Δϕ=0$ with the short pulse initial data $(ϕ,\partial_tϕ)(1,x)=\big(δ^{2-\varepsilon_0}ϕ_0(\frac{r-1}δ,ω), δ^{1-\varepsilon_0}ϕ_1(\frac{r-1}δ,ω)\big)$, where $p\in\mathbb N$, $p\geq 2$, $00$ is sufficiently small, under the outgoing constraint condition $(\partial_t+\partial_r)^kϕ(1,x)=O(δ^{2-\varepsilon_0})$ for $k=1,2$, we will establish the global existence of smooth large data solution $ϕ$ when $p>p_c$ with $p_c=\frac{1}{1-\varepsilon_0}$ being the critical exponent. In the forthcoming paper, when $1\leq p\leq p_c$, we show the formation of the outgoing shock before the time $t=2$ under the suitable assumptions of $(ϕ_0,ϕ_1)$.

Near-infrared II fluorescence imaging holds great promise for in vivo imaging and imaging-guided surgery with deep penetration and high spatiotemporal resolution. However, most NIR-II aromatic luminophores suffer from the notorious aggregation-caused quenching (ACQ) effect in the aqueous solution, which largely hinders their biomedical application in vivo. In this study, the first NIR-II organic aggregation-induced emission (AIE) fluorophore (HLZ-BTED), encapsulated as nanoparticles (HLZ-BTED dots) for in vivo biomedical imaging, was designed and synthesized. The NIR-II AIE HLZ-BTED dots showed high temporal resolution, high photostability, outstanding water-solubility and biocompatibility in vitro and in vivo. The HLZ-BTED dots were further used for long-term breast tumor imaging and visualizing tumor-feeding blood vessels, long-term hind limb vasculature and incomplete hind limb ischemia. More importantly, as a proof-of-concept, this is the first time that non-invasive and real-time NIR-II imaging of the gastrointestinal tract in health and disease has been performed, making the AIE dots a promising tool for gastrointestinal (GI) tract research, such as understanding the healthy status of GI peristalsis, diagnosing and evaluating intestinal motility dysfunction, and assessing drug effects on intestinal obstruction.

Solar cells with polysilicon-based passivating and carrier selective contacts are in the nascent stages of industrialization. Advances in commercially available metallization pastes are necessary to further improve the efficiency of industrial polysilicon-based solar cells and approach the best-reported lab efficiency of such cells. In this work, we have analyzed the front and rear screen-printed metallization of large-area n-type monoPolyTM cells with rear-side low-pressure chemical vapor deposited (LPCVD) polysilicon. Metallization pastes specifically optimized for contacting poly-Si layers are used in this work. The metallization is evaluated by studying its impact on the solar cell performance and by further electrical and optical characterization. An efficiency improvement of 0.5% absolute is demonstrated due to improvements in metallization pastes with a 23% champion cell efficiency (cell area: 244.3 cm2, busbarless metallization). An analysis of the fabricated cells indicates that the efficiency is currently limited by recombination at the front surface, both at the passivated and the metallized regions.

Novel biocompatible NIR-II aggregation-induced emission dots are facilely assembled and used for NIR-II biomedical fluorescence imaging., Near-infrared II fluorescence imaging holds great promise for in vivo imaging and imaging-guided surgery with deep penetration and high spatiotemporal resolution. However, most NIR-II aromatic luminophores suffer from the notorious aggregation-caused quenching (ACQ) effect in the aqueous solution, which largely hinders their biomedical application in vivo. In this study, the first NIR-II organic aggregation-induced emission (AIE) fluorophore (HLZ-BTED), encapsulated as nanoparticles (HLZ-BTED dots) for in vivo biomedical imaging, was designed and synthesized. The NIR-II AIE HLZ-BTED dots showed high temporal resolution, high photostability, outstanding water-solubility and biocompatibility in vitro and in vivo. The HLZ-BTED dots were further used for long-term breast tumor imaging and visualizing tumor-feeding blood vessels, long-term hind limb vasculature and incomplete hind limb ischemia. More importantly, as a proof-of-concept, this is the first time that non-invasive and real-time NIR-II imaging of the gastrointestinal tract in health and disease has been performed, making the AIE dots a promising tool for gastrointestinal (GI) tract research, such as understanding the healthy status of GI peristalsis, diagnosing and evaluating intestinal motility dysfunction, and assessing drug effects on intestinal obstruction.

This paper is concerned with the lifespan of and the blowup mechanism for smooth solutions to the 2-D nonlinear wave equation ∂ t 2 u − ∑ i = 1 2 ∂ i ( c i 2 ( u ) ∂ i u ) \partial _t^2u-\sum _{i=1}^2\partial _i(c_i^2(u)\partial _iu) = 0 =0 , where c i ( u ) ∈ C ∞ ( R n ) c_i(u)\in C^{\infty }(\mathbb {R}^n) , c i ( 0 ) ≠ 0 c_i(0)\neq 0 , and ( c 1 ′ ( 0 ) ) 2 + ( c 2 ′ ( 0 ) ) 2 ≠ 0 (c_1’(0))^2+(c_2’(0))^2\neq 0 . This equation has an interesting physical background as it arises from the pressure-gradient model in compressible fluid dynamics and also in nonlinear variational wave equations. Under the initial condition ( u ( 0 , x ) , ∂ t u ( 0 , x ) ) = ( ε u 0 ( x ) , ε u 1 ( x ) ) (u(0,x), \partial _tu(0,x))=(\varepsilon u_0(x), \varepsilon u_1(x)) with u 0 ( x ) , u 1 ( x ) ∈ C 0 ∞ ( R 2 ) u_0(x), u_1(x)\in C_0^{\infty }(\mathbb {R}^2) , and ε > 0 \varepsilon >0 is small, we will show that the classical solution u ( t , x ) u(t,x) stops to be smooth at some finite time T ε T_{\varepsilon } . Moreover, blowup occurs due to the formation of a singularity of the first-order derivatives ∇ t , x u ( t , x ) \nabla _{t,x}u(t,x) , while u ( t , x ) u(t,x) itself is continuous up to the blowup time T ε T_{\varepsilon } .

For the three-dimensional (3D) quasilinear wave equation $\sum_{i,j=0}^3g^{ij}(\partial u)\partial_{ij}^2u=0$ with coefficients independent of the solution $u$, a blowup result or global existence result has been completely established provided that the null condition does not hold or holds, respectively. In this paper and a subsequent paper (see [D. Bingbing, W. Ingo, and Y. Huicheng, The Small Data Solutions of General $3$-D Quasilinear Wave Equations. II, preprint, 2014]), we will systematically study the more general 3D quasilinear wave equation $\sum_{i,j=0}^3g^{ij}(u, \partial u)\partial_{ij}^2u=0$ with coefficients depending simultaneously on $u$ and $\partial u$. When the weak null condition holds, we show that the smooth small data solution exists globally in the paper.

The 3-D pressure-gradient system arises from the splitting of the three-dimensional compressible Euler system, which can be reduced into such a nonlinear wave equation ∂t2v−div(ev∇v)=0, where div(ev∇v)=∑i=13∂i(ev∂iv), ∂i=∂xi (1⩽i⩽3) and x=(x1,x2,x3). In this paper, we will establish a blowup result of classical solutions to the 3-D pressure-gradient systems when the initial data are of small perturbations with respect to constant states.

For the 2-D quasilinear wave equation $\displaystyle \sum_{i,j=0}^2g_{ij}(\nabla u)\partial_{ij}u=0$ with coefficients independent of the solution $u$, a blowup result for small data solutions has been established in [1,2] provided that the null condition does not hold and a generic nondegeneracy condition on the initial data is fulfilled. In this paper, we are concerned with the more general 2-D quasilinear wave equation $\displaystyle \sum_{i,j=0}^2g_{ij}(u, \nabla u)\partial_{ij}u=0$ with coefficients that depend simultaneously on $u$ and $\nabla u$. When the null condition does not hold and a suitable nondegeneracy condition on the initial data is satisfied, we show that smooth small data solutions blow up in finite time. Furthermore, we derive an explicit expression for the lifespan and establish the blowup mechanism., 27 pages

This paper is concerned with the small smooth data problem for the 3-D nonlinear wave equation $\partial_t^2u-\left (1+u+\p_t u\right)\Delta u=0$. This equation is prototypical of the more general equation $\dsize\sum_{i,j=0}^3g_{ij}(u, \nabla u)\partial_{ij}u=0$, where $x_0=t$ and $g_{ij}(u, \nabla u)=c_{ij}+d_{ij}u+\dsize\sum_{k=0}^3e_{ij}^k\partial_ku+O(|u|^2+|\nabla u|^2)$ are smooth functions of their arguments, with $c_{ij}, d_{ij}$ and $e_{ij}^k$ being constants, and $d_{ij}\neq0$ for some $(i,j)$; moreover, $\dsize\sum_{i,j,k=0}^3e_{ij}^k(\partial_ku)\p_{ij} u$ does not fulfill the null condition. For the 3-D nonlinear wave equations $\partial_t^2u-\left (1+u\right)\Delta u=0$ and $\partial_t^2u-\left (1+\partial_t u\right)\Delta u=0$, H. Lindblad, S. Alinhac, and F. John proved and disproved, respectively, the global existence of small smooth data solutions. For radial initial data, we show that the small smooth data solution of $\partial_t^2u-\left(1+u+\partial_t u\right)\Delta u=0$ blows up in finite time. The explicit expression of the asymptotic lifespan $T_{\varepsilon}$ as $\varepsilon\to0^+$ is also given., Comment: 20 pages, 2 figures

This paper is concerned with the lifespan and the blowup mechanism for smooth solutions to the 2-D nonlinear wave equation $\p_t^2u-\ds\sum_{i=1}^2\p_i(c_i^2(u)\p_iu)$ $=0$, where $c_i(u)\in C^{\infty}(\Bbb R^n)$, $c_i(0)\neq 0$, and $(c_1'(0))^2+(c_2'(0))^2\neq 0$. This equation has an interesting physics background as it arises from the pressure-gradient model in compressible fluid dynamics and also in nonlinear variational wave equations. Under the initial condition $(u(0,x), \p_tu(0,x))=(\ve u_0(x), \ve u_1(x))$ with $u_0(x), u_1(x)\in C_0^{\infty}(\Bbb R^2)$, and $\ve>0$ is small, we will show that the classical solution $u(t,x)$ stops to be smooth at some finite time $T_{\ve}$. Moreover, blowup occurs due to the formation of a singularity of the first-order derivatives $\na_{t,x}u(t,x)$, while $u(t,x)$ itself is continuous up to the blowup time $T_{\ve}$., 22 pages