Theme relevancy judging method and device

The invention provides a theme relevancy judgment method and device. The method comprises the steps: constructing a webpage feature vector for an obtained webpage; calculating the similarity between the selected topic feature vector and the webpage feature vector by using a pre-trained semantic vector space model; and screening out the webpage feature vectors of which the similarity is higher thana preset value. According to the method and device, the advantages of semantic vector similarity calculation and a machine learning method are combined, compared with the prior art, high judgment precision can be achieved, and improvement different from the prior art is made on the aspect of training sample screening. 本申请提供了一种主题相关度判别方法及装置，方法包括：对获取的网页构建网页特征向量；利用预先训练的语义向量空间模型对选定的主题特征向量与网页特征向量之间的相似度进行计算；筛选出相似度高于预设值的网页特征向量。本申请结合了语义向量相似度计算和机器学习方法的优点，相比于现有技术，可以实现较高的判别精度，并且本申请在训练样本的筛选上也作出了不同于现有技术的改进。