Semantic textual similarity(STS) is a common task in natural language processing(NLP). STS measures the degree of semantic equivalence of two textual snippets. Recently, machine learning methods have been applied to t...Semantic textual similarity(STS) is a common task in natural language processing(NLP). STS measures the degree of semantic equivalence of two textual snippets. Recently, machine learning methods have been applied to this task, including methods based on support vector regression(SVR). However, there exist amounts of features involved in the learning process, part of which are noisy features and irrelative to the result.Furthermore, different parameters will significantly influence the prediction performance of the SVR model. In this paper, we propose genetic algorithm(GA) to select the effective features and optimize the parameters in the learning process, simultaneously. To evaluate the proposed approach, we adopt the STS-2012 dataset in the experiment. Compared with the grid search, the proposed GA-based approach has better regression performance.展开更多
目的全景图像的质量评价和传输、处理过程并不是在同一个空间进行的,传统的评价算法无法准确地反映用户在观察球面场景时产生的真实感受,针对观察空间与处理空间不一致的问题,本文提出一种基于相位一致性的全参考全景图像质量评价模型...目的全景图像的质量评价和传输、处理过程并不是在同一个空间进行的,传统的评价算法无法准确地反映用户在观察球面场景时产生的真实感受,针对观察空间与处理空间不一致的问题,本文提出一种基于相位一致性的全参考全景图像质量评价模型。方法将平面图像进行全景加权,使得平面上的特征能准确反映球面空间质量畸变。采用相位一致性互信息的相似度获取参考图像和失真图像的结构相似度。接着,利用相位一致性局部熵的相似度反映参考图像和失真图像的纹理相似度。将两部分相似度融合可得全景图像的客观质量分数。结果实验在全景质量评价数据集OIQA(omnidirectional image quality assessment)上进行,在原始图像中引入4种不同类型的失真,将提出的算法与6种主流算法进行性能对比,比较了基于相位信息的一致性互信息和一致性局部熵,以及评价标准依据4项指标。实验结果表明,相比于现有的6种全景图像质量评估算法,该算法在PLCC(Pearson linear correlation coefficient)和SRCC(Spearman rank order correlation coefficient)指标上比WS-SSIM(weighted-to-spherically-uniform structural similarity)算法高出0.4左右,并且在RMSE(root of mean square error)上低0.9左右,4项指标最优,能够获得更好的拟合效果。结论本文算法解决了观察空间和映射空间不一致的问题,并且融合了基于人眼感知的多尺度互信息相似度和局部熵相似度,获得与人眼感知更为一致的客观分数,评价效果更为准确,更加符合人眼视觉特征。展开更多
文摘Semantic textual similarity(STS) is a common task in natural language processing(NLP). STS measures the degree of semantic equivalence of two textual snippets. Recently, machine learning methods have been applied to this task, including methods based on support vector regression(SVR). However, there exist amounts of features involved in the learning process, part of which are noisy features and irrelative to the result.Furthermore, different parameters will significantly influence the prediction performance of the SVR model. In this paper, we propose genetic algorithm(GA) to select the effective features and optimize the parameters in the learning process, simultaneously. To evaluate the proposed approach, we adopt the STS-2012 dataset in the experiment. Compared with the grid search, the proposed GA-based approach has better regression performance.
文摘目的全景图像的质量评价和传输、处理过程并不是在同一个空间进行的,传统的评价算法无法准确地反映用户在观察球面场景时产生的真实感受,针对观察空间与处理空间不一致的问题,本文提出一种基于相位一致性的全参考全景图像质量评价模型。方法将平面图像进行全景加权,使得平面上的特征能准确反映球面空间质量畸变。采用相位一致性互信息的相似度获取参考图像和失真图像的结构相似度。接着,利用相位一致性局部熵的相似度反映参考图像和失真图像的纹理相似度。将两部分相似度融合可得全景图像的客观质量分数。结果实验在全景质量评价数据集OIQA(omnidirectional image quality assessment)上进行,在原始图像中引入4种不同类型的失真,将提出的算法与6种主流算法进行性能对比,比较了基于相位信息的一致性互信息和一致性局部熵,以及评价标准依据4项指标。实验结果表明,相比于现有的6种全景图像质量评估算法,该算法在PLCC(Pearson linear correlation coefficient)和SRCC(Spearman rank order correlation coefficient)指标上比WS-SSIM(weighted-to-spherically-uniform structural similarity)算法高出0.4左右,并且在RMSE(root of mean square error)上低0.9左右,4项指标最优,能够获得更好的拟合效果。结论本文算法解决了观察空间和映射空间不一致的问题,并且融合了基于人眼感知的多尺度互信息相似度和局部熵相似度,获得与人眼感知更为一致的客观分数,评价效果更为准确,更加符合人眼视觉特征。
