Quantum-Inspired Complex-Valued Language Models for Aspect-Based Sentiment Classification

Aiming at classifying the polarities over aspects, aspect-based sentiment analysis (ABSA) is a fine-grained task of sentiment analysis. The vector representations of current models are generally constrained to real values. Based on mathematical formulations of quantum theory, quantum language models have drawn increasing attention. Words in such models can be projected as physical particles in quantum systems, and naturally represented by representation-rich complex-valued vectors in a Hilbert Space, rather than real-valued ones. In this paper, the Hilbert Space representation for ABSA models is investigated and the complexification of three strong real-valued baselines are constructed. Experimental results demonstrate the effectiveness of complexification and the outperformance of our complex-valued models, illustrating that the complex-valued embedding can carry additional information beyond the real embedding. Especially, a complex-valued RoBERTa model outperforms or approaches the previous state-of-the-art on three standard benchmarking datasets.     

Assessment of Chemical Composition and Anti-Penicillium Activity of Vapours of Essential Oils from Abies Alba and Two Melaleuca Species in Food Model Systems

The possibilities of the practical utilization of essential oils (EOs) from various plant species in the food industry have attracted the attention of the scientific community. Following our previous studies, the antifungal activities of three further commercial EOs, Melaleuca armillaris subsp. armillaris (rosalina; REO), Melaleuca quinquenervia (niaouli; NEO), and Abies alba (fir; FEO), were evaluated in the present research in respect to their chemical profiles, over four different concentrations, 62.5 μL/L, 125 μL/L, 250 μL/L, and 500 μL/L. The findings revealed that the major compounds of REO, NEO, and FEO were linalool (47.5%), 1,8-cineole (40.8%), and α-pinene (25.2%), respectively. In vitro antifungal determinations showed that the inhibition zones of a Penicillium spp. mycelial growth ranged from no inhibitory effectiveness (00.00 ± 00.00 mm) to 16.00 ± 1.00 mm, indicating a very strong antifungal activity which was detected against P. citrinum after the highest REO concentration exposure. Furthermore, the in situ antifungal efficacy of all EOs investigated was shown to be dose-dependent. In this sense, we have found that the highest concentration (500 µL/L) of REO, NEO, and FEO significantly reduced (p < 0.05) the growth of all Penicillium strains inoculated on the bread, carrot, and potato models. These results indicate that the investigated EOs may be promising innovative agents in order to extend the shelf life of different types of food products, such as bread, carrot and potato.     

圆柱类构件超声相控阵聚焦模型

在利用平面相控阵超声换能器对圆柱类构件检测时,受曲界面结构引起的入射波和回波时延的影响,扫描声束的波阵面产生弯曲,在利用传统迭代遍历算法计算延迟时间时效率低,无法发挥相控阵换能器检测优势。针对上述问题,建立了一种基于圆柱类构件特征和耦合介质特性的超声相控阵扫描成像的聚焦模型。该模型基于换能器、耦合介质、圆柱类构件材料特性和几何关系以及声线模型和折射定律,建立了耦合介质及被检构件的声速、曲面曲率半径、阵列与曲面间的距离等关联的延迟时间聚焦控制模型。利用该模型计算出的延时时间,对各阵元发射时间进行控制,从而实现扫描声束的聚焦。该文以液体和有机玻璃制介质楔块为例,对圆柱钢曲面的相控阵声束聚焦进行了仿真实验,结果表明该文方法计算效率显著提升,同时声束可以在预设位置实现聚焦,验证了该模型在计算效率上的优势与有效性。     

Interplay Between Quark-Antiquark and Diquark Condensates in Vacuum in a Two-Flavor Nambu-Jona-Lasinio Model

By means of a relativistic effective potential, we analytically research competition between the quark- antiquark condensates （qq） and the diquark condensates （qq） in vacuum in ground state of a two-flavor Nambu Jona Lasinio （NJL） model and obtain the Gs-Hs phase diagram, where Gs and Hs are the respective four-fermion coupling constants in scalar quark-antiquark channel and scalar color anti-triplet diquark channel. The results show that, in the chiral limit, there is only the pure （qq） phase when Gs/Hs 〉 2/3, and as Gs/Hs decreases to 2/3 〉 Gs/Hs ≥ 0 one will first have a coexistence phase of the condensates （qq） and （qq） and then a pure （qq） phase. In non-zero bare quark mass case, the critical value of Gs/Hs at which the pure （qq） phase will transfer to the coexistence phase of the condensates （qq） and （qq） will be less than 2/3. Our theoretical results, combined with present phenomenological fact that there is no diquark condensates in the vacuum of QCD, will also impose a real restriction to any given two-flavor NJL model which is intended to simulate QCD, i.e. in such model the resulting sma/lest ratio Gs/Hs after the Fierz transformations in the Hartree approximation must be larger than 2/3. A few phenomenological QCD-like NJL models are checked and analyzed.     

Positions and Widths of Anticrossings for Potassium Rydberg Stark States

The B-spline expansion technique is applied to study the anticrossings for potassium Rydberg states in a static electric field. The results of our calculation indicate that the antierossings are caused mainly by the core interaction or by the fine structure interaction. Our results for the positions and the widths of the anticrossings are in good agreement with experimental data.     

偏序集上的滤子极大理想

在偏序集上引入并考察了滤子极大理想的概念,证明了相应的存在性定理。引入并考察了伪极大元和伪既约元的概念,利用图表的形式对连续格中各种类型的既约元和素元之间的关系进行了归纳总结,完善了文献《Continuous Lattices and Domains》(作者:G.Gierz,et al)中的一个图表的相关内容,填补了在分配的连续格情形该图表的一个未知内容,部分地回答了该文献中的一个问题。     

基于改进GMM与帧差法的运动棉杂率分析算法研究

针对清梳棉流程中产出棉含杂高、质量差的问题,结合改进高斯混合模型(Gaussian mixed model,GMM)与帧差法,提出一种局部运动棉杂率控制优化方法。首先针对清棉机除杂原理及棉杂特性进行分析;其次通过提取视频关键帧并改进GMM与帧差法对图像序列“与”运算实现目标的精确提取,进而通过设计GMM分类器获得棉杂率并进行分析;最后与传统的检测算法作对比验证。实验表明,改进后的算法在有效性以及实用性方面优于传统算法。同时,通过引入闭环控制能满足工业高精度、实时性的需求。     

融合注意力机制的轻量级摔倒检测

摔倒检测大都依靠传感器设备,此类方法受设备自身和环境因素影响较大,常常无法发挥该有的作用,同时,基于视觉的方法往往实时性较差,鲁棒性不强。针对上述问题,本文提出了一种鲁棒性强、能有效部署在嵌入式设备上的轻量化摔倒检测算法。以YOLOv5为基准模型,首先,融合轻量级注意力机制模块,使网络更关注要识别的目标区域,增强网络的识别精度。其次,使用模型压缩方法对模型进行剪枝,减小模型体积和计算量,使模型轻量化,以提高推理速度和便于部署在嵌入式设备中。最后,对剪枝后的模型进行知识蒸馏,在不提升模型复杂度的前提下提升模型的检测精度。实验结果表明：本文模型相较于基准模型,mAP增加了1.7%,召回率提高了1.2%,模型体积减小了79.1%,浮点运算量降低了70.9%。将本文模型部署在嵌入式设备Jetson Nano上,检测速率达到13.2 frame/s,基本满足实时性摔倒检测的要求。     

Ordinal Regression for Information Retrieval

This letter presents a new discriminative model for Information Retrieval （IR）, referred to as Ordinal Regression Model （ORM）. ORM is different from most existing models in that it views IR as ordinal regression problem （i.e. ranking problem） instead of binary classification. It is noted that the task of IR is to rank documents according to the user information needed, so IR can be viewed as ordinal regression problem. Two parameter learning algorithms for ORM are presented. One is a perceptron-based algorithm. The other is the ranking Support Vector Machine （SVM）. The effectiveness of the proposed approach has been evaluated on the task of ad hoc retrieval using three English Text REtrieval Conference （TREC） sets and two Chinese TREC sets. Results show that ORM significantly outperforms the state-of-the-art language model approaches and OKAPI system in all test sets; and it is more appropriate to view IR as ordinal regression other than binary classification.     

用户初始服务密码管理方式

本文对用户初始服务密码管理困难的原因进行了深入的分析,提出了将初始服务密码印制在SIM卡涂层下的新型管理方式。