基于跨场景推理的事件关系检测方法

事件关系检测是一项面向事件之间逻辑关系的自然语言处理技术。事件关系识别的核心任务是以事件为基本语义单元,通过分析事件的篇章结构信息及语义特征,实现事件逻辑关系的深层检测。该文首次建立一套事件关系检测的任务和研究体系,包括任务定义、关系体系划分、语料采集与标注、评价方法等。同时,该文提出了一种跨场景推理的事件关系检测方法,该方法认为,具有相同事件场景的“事件对”,往往具有相同的事件关系类型。该文提出的基于跨场景推理的事件关系检测方法在针对四大类事件关系类型的检测精确率为54.21%。     

基于句法语义特征的中文实体关系抽取

实体关系抽取的核心问题是实体关系特征的选择。以往的研究通常都以词法特征、实体原始特征等来刻画实体关系,其抽取效果已难再提高。在传统方法的基础上,该文提出一种基于句法特征、语义特征的实体关系抽取方法,融入了依存句法关系、核心谓词、语义角色标注等特征,选择SVM作为机器学习的实现途径,以真实新闻文本作为语料进行实验。实验结果表明该方法的F1值有明显提升。     

情感分析中极性副词的自动扩展

为了扩展尽可能多的极性副词（基于词典WordNet 3. 0）, 提出了三种类型的查询字符串。除了WordNet中的同义关系和反义关系, 还使用了近义（similar to）和又见（also see）这两种关系。实验数据表明, 后两种关系比前两种关系在扩展极性副词方面更重要。详细的实验验证了所提出的POAE算法加Transform转换函数在自动扩展极性副词方面是可行的。     

The study on the conductivity of styrene-maleic anhydride copolymer electrolytes based on poly(ethylene glycol)400 and LiClO4

Polyelectrolytes, in this study were synthesized from styrene-maleic anhydride (SMA) copolymer, poly(ethylene glycol)400 (PEG400), and lithium perchlorate (LiClO₄). Fourier transform infrared spectroscopy (FTIR), and magic angle spinning (MAS) solid-state NMR were used to monitor the interaction between Li⁺ ions and polymer. The results of FTIR and MAS solid-state NMR indicate the Li⁺ ions are preferentially coordinated to the ether oxygen of PEG. The T_g of the PEG segments in polyelectrolyte increases with LiClO₄ concentration, as determined by differential scanning calorimetry (DSC), indicating that solubility of the Li⁺ ions in the host polymer increases with the PEG content. Impedance spectroscopy (IS) shows that the bulk conductivity of polyelectrolytes and the conductivity behavior obeys the Vogel-Tamman-Fulcher (VTF) equation.     

基本主方向关系的反关系推理

在深入研究了基于MBR的主方向关系的反关系推理的基础上,提出了一种基于区域对象本身的基本主方向关系的反关系推理算法。在理论上对该算法的正确性和完备性进行了证明,并通过与实际情形逐一对比验证了该算法的正确性。     

Assessment of a redox alkaline/iron-chelate absorption process for the removal of dilute hydrogen sulfide in air emissions

The oxidative absorption of hydrogen sulfide (H₂S) into a solution of ferric chelate of trans-1,2- diaminocyclohexanetetraacetate (CDTA) was studied in a counter-current laboratory column randomly packed with 15 mm plastic Ralu rings. The present investigation takes concern about the Kraft pulping situation where dilute H₂S concentrations are omnipresent in large-volume gas effluents. A fractional two-level factorial approach was instigated to determine the significance of six operating variables, namely the solution's alkalinity (pH; 8.5-10.5), the liquid mass flow rate (L;1.73-), the solution's ionic strength (I_C;0.01-), the gas mass flow rate (G;0.19-), the inlet H₂S concentration (C_H2S,0;70-430 ppm) and the initial ferric CDTA concentration (C_Fe,0;100 -). Initially, a Plackett-Burman design matrix of seven duplicated experiments revealed that pH is the leading factor controlling the H₂S conversion rate while the ionic strength and ferric CDTA concentration effects remained negligible within the factorial domain. Surface response analysis based on 11 duplicated factorial experiments plus 10 central composite trials revealed that the H₂S conversion significantly increases with liquid flow rate but decreases with growing H₂S load up. Further examination about the influence of ferric CDTA on H₂S absorption rate was set up over a broader concentration range (C_Fe,0;0- at pH of 9.5 and 10.5. It showed good potential at as H₂S conversion increased by a significant 25% for both pH values in comparison to pure alkaline solutions containing no ferric CDTA.     

基于本体视图特征项抽取方法研究*

为提供比单纯词汇信息更高效的概念特征信息和深层语义信息,并满足面向同一文本的多检索需求,在半自动化智能检索框架中引入本体视图,提出一种基于本体视图的特征项抽取方法。此方法首先针对文本特征建立本体视图;然后结合文本信息进行特征项抽取和类型映射,得到特征项集;最后基于特征项集完成检索过程。检索结果显示,基于本体视图特征项抽取方法能改善检索系统的性能,提高检索的准确率和效率。