氯乙烯变温吸附装置的研究

介绍了采用变温吸附技术对氯乙烯进行干燥脱水的工艺流程及运行效果,阐述了变温吸附装置运行过程中出现的问题及相应的解决方法。     

Sequential alignment attention model for scene text recognition

Scene text recognition has been a hot research topic in computer vision due to its various applications. The state-of-the-art solutions usually depend on the attention-based encoder-decoder framework that learns the mapping between input images and output sequences in a purely data-driven way. Unfortunately, there often exists severe misalignment between feature areas and text labels in real-world scenarios. To address this problem, this paper proposes a sequential alignment attention model to enhance the alignment between input images and output character sequences. In this model, an attention gated recurrent unit (AGRU) is first devised to distinguish the text and background regions, and further extract the localized features focusing on sequential text regions. Furthermore, CTC guided decoding strategy is integrated into the popular attention-based decoder, which not only helps to boost the convergence of the training but also enhances the well-aligned sequence recognition. Extensive experiments on various benchmarks, including the IIIT5k, SVT, and ICDAR datasets, show that our method substantially outperforms the state-of-the-art methods.     

Investigation on thermal conductivity and mechanical properties of Si3N4 ceramics via one-step sintering

High thermal conductivity Si₃N₄ ceramics were fabricated using a one-step method consisting of reaction-bonded Si₃N₄ (RBSN) and post-sintering. The influence of Si content on nitridation rate, β/(α+β) phase rate, thermal conductivity and mechanical properties was investigated in this work. It is of special interest to note that the thermal conductivity showed a tendency to increase first and then decrease with increasing Si content. This experimental result shows that the optimal thermal conductivity and fracture toughness were obtained to be 66 W (m K)^-1 and 12.0 MPa m^1/2, respectively. As a comparison, the nitridation rate and β/(α+β) phase rate in a static pressure nitriding system, i.e., 97% (MS10), 97% (MS15), 97% (MS20) and 8.3% (MS10), 8.3% (MS15), 8.9% (MS20), respectively, have obvious advantages over those in a flowing nitriding system, i.e., 91% (MS10), 91% (MS15), 93% (MS20) and 3.1% (MS10), 3.3% (MS15), 3.3% (MS20), respectively. Moreover, high lattice integrity of the β-Si₃N₄ phase was observed, which can effectively confine O atoms into the β-Si₃N₄ lattice using MgO as a sintering additive. This result indicates that one-step sintering can provide a new route to prepare Si₃N₄ ceramics with a good combination of thermal conductivity and mechanical properties.     

Optimizing the stability and electrical transport properties of CeNbO4+δ-based oxide ceramics by regulating oxygen ion content

CeNbO_4+δ ceramics have attracted extensive research interest because of their unique mixed ion-electron transport characteristics and interesting structure-functional characteristics caused by the difference in oxygen ion content. Although the change of oxygen ion content brings rich redox properties, it also causes serious crystal transformation and abnormal electrical transport properties. In order to obtain stable structure and excellent electrical transport properties, the directional regulation of the oxygen ion content has been realized through introducing Al₂O₃ and high temperature aging. After 600 h of aging at 1073 K, the prepared composite ceramics not only obtain a stable structure without crystal transformation, but also show good negative temperature coefficient (NTC) thermistor characteristics in the temperature range of 473 K–1273 K, in which the linear fitting maximum Pearson's r of the relationship between lnρ and 1000/T can reach 99.97%. The proposed method provides a new thought for the design and application of high-temperature electronic ceramics.     

卡尔费休法测定液态氯乙烯中水含量的研究

采用卡尔费休法测定了液态氯乙烯单体中的水含量。试验确定了采样工具、样品的预处理方式及选用溶剂的最优方案,并进行了精密度与准确度的验证。结果表明卡尔费休法测定液态氯乙烯单体中的水含量准确、快速,可以用来准确、及时地指导生产。     

氮含量对HPD-1双相不锈钢组织及力学性能的影响

通过显微组织观察与力学性能测试研究了氮含量(0.08%~0.22％,质量分数)对HPD-1双相不锈钢硬度、拉伸性能、低温冲击性能及疲劳性能的影响。结果表明,氮含量变化可显著影响试验钢γ/α相比例,当氮含量由0.08％升高到0.22％,γ相含量由38.1％提高至56.5％。α相的硬度由273 HV10提高到343 HV10,γ相的硬度由238 HV10提高到299 HV10,试验钢强度明显提升。氮元素对两相比例和奥氏体相韧性的双重影响导致试验钢低温冲击性能呈先上升后下降的趋势;更高的氮含量抑制疲劳裂纹萌生与拓展,是影响HPD-1双相不锈钢室温疲劳性能的主要因素。撕裂棱是疲劳断口的显著特征。     

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。     

面向智能客服系统的情感分析技术

该文以阿里小蜜为例,对智能客服系统中的情感分析技术进行比较全面的介绍,包括情感分析算法模型的原理及其在智能客服系统的多个应用场景中的实际落地使用方式和效果分析。智能客服在解决客户高频业务问题的同时,也需要给客户提供多维度的、具有类人能力的助理、导购、语聊和娱乐等服务,提高客户对智能客服机器人的整体满意度。在此过程中,情感分析技术在机器人类人能力建设中起到了至关重要的作用。该文围绕智能客服系统中人机结合的服务形式,从六个维度总结和介绍了情感分析技术在智能客服系统中的应用场景,包括用户情感检测、用户情感安抚、情感生成式语聊、客服服务质检、会话满意度预估和智能人工入口。     

基于语言特征自动获取的反问句识别方法

反问句是以疑问的形式表达强烈情感的修辞方式,对其有效识别可为自然语言处理中的情感分析任务提供技术支持。该文提出了一种基于语言特征自动获取的反问句识别方法。首先,利用标签注意机制,建立了一个数据驱动的特征抽取模型,用于获取与任务相关的词汇、句法结构、符号标记和话题等语言特征。其次,利用Bi-LSTM模型分别对句子和语言特征进行表示,两者的交互注意被用于获取句子的各个词和符号的注意力权重向量。该权重向量作用于句子的表示,用于构建一个强化语言特征的反问句识别模型。在中文微博数据集上的实验结果表明,提出的方法与之前的工作相比,反问句识别性能有显著提升。