In order to judge whether the user reviews are relevant to App software, this paper proposed a method to judge the relevance of user reviews based on Naive Bayesian text classification and term frequency.Firstly, the ...In order to judge whether the user reviews are relevant to App software, this paper proposed a method to judge the relevance of user reviews based on Naive Bayesian text classification and term frequency.Firstly, the keywords sets of App software’s user reviews are extracted. Then, the keywords sets are optimized. Finally, the relevance score of the user reviews are calculated, and whether the user reviews are relevant is judged. Through the experiment, this method is proved that can judge the relevance of App software’s user reviews effectively.展开更多
With the relevance theory as a guideline and Luoyang's tourism texts as an example, this paper uses the method of exemplification and comparison, aims to explore English translation problems of Luoyang's touri...With the relevance theory as a guideline and Luoyang's tourism texts as an example, this paper uses the method of exemplification and comparison, aims to explore English translation problems of Luoyang's tourism texts, and puts forward practical translation strategies respectively, thus helping to solve the problems in the English translation of tourism texts and promoting Chinese influence in the international context.展开更多
为充分利用MOOC(massive open online course)上下文信息,精确表示学习者和课程特征,提出一种多特征融合的MOOC推荐模型(multi-feature fusion based model for MOOC recommendation,MFF-MOOCREC)。利用文本卷积神经网络和双向长短时记...为充分利用MOOC(massive open online course)上下文信息,精确表示学习者和课程特征,提出一种多特征融合的MOOC推荐模型(multi-feature fusion based model for MOOC recommendation,MFF-MOOCREC)。利用文本卷积神经网络和双向长短时记忆网络捕获数据中的文本和时序特征,并设计多级注意力机制提取学习者交互序列、评论文本和课程多元属性中的关键信息;基于前缀投影的模式挖掘和亲和力传播算法对原始课程类别进行关联聚类分析以增加推荐的覆盖率;采用概率矩阵分解训练模型参数,将优化后的学习者隐向量和课程隐向量点乘产生预测评分。实验表明,和现有推荐方法相比,MFF-MOOCREC的命中率、归一化折损累计增益和覆盖率指标在Coursera数据集上平均提高46.86%、41.19%和10.95%,在iCourse数据集上平均提高44.08%、28.79%和9.81%,对于缓解数据稀疏问题,提升推荐质量具有一定优势。展开更多
锚文本作为对目标网页的描述,往往分布在不同的源网页上,质量也参差不齐.本文利用了超链接分析算法的成果,提出一种基于源网页质量的锚文本相似度计算方法--LAAT(Link Aid Anchor Text).实验表明,利用源网页质量能够有效地综合各源网页...锚文本作为对目标网页的描述,往往分布在不同的源网页上,质量也参差不齐.本文利用了超链接分析算法的成果,提出一种基于源网页质量的锚文本相似度计算方法--LAAT(Link Aid Anchor Text).实验表明,利用源网页质量能够有效地综合各源网页上的锚文本组成,从而能够提高检索性能.展开更多
文摘In order to judge whether the user reviews are relevant to App software, this paper proposed a method to judge the relevance of user reviews based on Naive Bayesian text classification and term frequency.Firstly, the keywords sets of App software’s user reviews are extracted. Then, the keywords sets are optimized. Finally, the relevance score of the user reviews are calculated, and whether the user reviews are relevant is judged. Through the experiment, this method is proved that can judge the relevance of App software’s user reviews effectively.
文摘With the relevance theory as a guideline and Luoyang's tourism texts as an example, this paper uses the method of exemplification and comparison, aims to explore English translation problems of Luoyang's tourism texts, and puts forward practical translation strategies respectively, thus helping to solve the problems in the English translation of tourism texts and promoting Chinese influence in the international context.
文摘为充分利用MOOC(massive open online course)上下文信息,精确表示学习者和课程特征,提出一种多特征融合的MOOC推荐模型(multi-feature fusion based model for MOOC recommendation,MFF-MOOCREC)。利用文本卷积神经网络和双向长短时记忆网络捕获数据中的文本和时序特征,并设计多级注意力机制提取学习者交互序列、评论文本和课程多元属性中的关键信息;基于前缀投影的模式挖掘和亲和力传播算法对原始课程类别进行关联聚类分析以增加推荐的覆盖率;采用概率矩阵分解训练模型参数,将优化后的学习者隐向量和课程隐向量点乘产生预测评分。实验表明,和现有推荐方法相比,MFF-MOOCREC的命中率、归一化折损累计增益和覆盖率指标在Coursera数据集上平均提高46.86%、41.19%和10.95%,在iCourse数据集上平均提高44.08%、28.79%和9.81%,对于缓解数据稀疏问题,提升推荐质量具有一定优势。