For a special use a new modelling method of evaluating external disturbing potential is presented in this paper.Being different from classical methods in physical geodesy this method is grounded upon the theory of uni...For a special use a new modelling method of evaluating external disturbing potential is presented in this paper.Being different from classical methods in physical geodesy this method is grounded upon the theory of unified representation of gravitational field.The models created in this way are particularly satisfactory for a high_speed computation of gravitational field in low altitude because they take account of topographic effects and have their kernel functions with simple structure and weak singularity.展开更多
大量有效样本标注是有监督学习性能的重要保证,但又存在耗时且人力成本高的问题.加之,在实际应用环境,很难在每个应用领域都有足够的标定样本数据支持分类器的训练.而将源领域所获的训练模型直接用于目标领域,又由于目标领域和源领域信...大量有效样本标注是有监督学习性能的重要保证,但又存在耗时且人力成本高的问题.加之,在实际应用环境,很难在每个应用领域都有足够的标定样本数据支持分类器的训练.而将源领域所获的训练模型直接用于目标领域,又由于目标领域和源领域信息分布差异,会导致跨领域分类器应用准确率降低的问题.针对以上问题,提出一种基于多视角共享特征的领域空间对齐的跨领域情感分类(domain alignment based on multi-viewpoint domain-shared feature for cross-domain sentiment classification,DAMF)算法.该算法首先通过融合多个情感词典,消除通过互信息值所选择的领域共享特征中情感词的极性分歧问题.在此基础上,以领域间无歧义共享特征为桥梁,结合通过语法规则提取的各领域中有相同极性的情感词对和通过关联规则学习的各领域中有强关联关系的特征词对,进行领域间相同极性的专有情感词对和强关联关系的特征词对的提取,构建目标领域和源领域数据的统一特征表示空间,减小了领域间因极性分歧和特征分布不同造成的差异,实现不同领域空间对齐.同时在公共数据集上的跨领域实验表明,基于多视角共享特征的领域空间对齐跨领域倾向性分析算法一定程度上提高了跨领域情感分类的准确率.展开更多
文摘For a special use a new modelling method of evaluating external disturbing potential is presented in this paper.Being different from classical methods in physical geodesy this method is grounded upon the theory of unified representation of gravitational field.The models created in this way are particularly satisfactory for a high_speed computation of gravitational field in low altitude because they take account of topographic effects and have their kernel functions with simple structure and weak singularity.
文摘大量有效样本标注是有监督学习性能的重要保证,但又存在耗时且人力成本高的问题.加之,在实际应用环境,很难在每个应用领域都有足够的标定样本数据支持分类器的训练.而将源领域所获的训练模型直接用于目标领域,又由于目标领域和源领域信息分布差异,会导致跨领域分类器应用准确率降低的问题.针对以上问题,提出一种基于多视角共享特征的领域空间对齐的跨领域情感分类(domain alignment based on multi-viewpoint domain-shared feature for cross-domain sentiment classification,DAMF)算法.该算法首先通过融合多个情感词典,消除通过互信息值所选择的领域共享特征中情感词的极性分歧问题.在此基础上,以领域间无歧义共享特征为桥梁,结合通过语法规则提取的各领域中有相同极性的情感词对和通过关联规则学习的各领域中有强关联关系的特征词对,进行领域间相同极性的专有情感词对和强关联关系的特征词对的提取,构建目标领域和源领域数据的统一特征表示空间,减小了领域间因极性分歧和特征分布不同造成的差异,实现不同领域空间对齐.同时在公共数据集上的跨领域实验表明,基于多视角共享特征的领域空间对齐跨领域倾向性分析算法一定程度上提高了跨领域情感分类的准确率.
