基于张量递归神经网络的英文语义关系分类方法研究

Research on the Classification of English Semantic Relationships Based on Tensor Recursive Neural Network

语义关系分类作为当前语义技术的一个基础领域,获得广泛的关注。提出基于张量空间的递归神经网络算法,利用张量(向量-矩阵对)表示单词,获得更准确的语义分类结果。通过无监督的结构化方式训练模型,大大简便分类过程,舍弃了人工手动标注。实验表明,该算法可以有效识别语义关系,比传统算法性能提高5%以上。

Classification of semantic relationships is a basic area of semantic technology and gains wide attention. Introduces a better approach to classify semantic relationships of words, tensor recursive neural network model which uses tensor （vector-matrix pairs）to represent a sin- gle words. The model trains the data by an unsupervised and structural way, which has no more need of hand-labeled corpus and simplify the process of classification. The experiment shows that the algorithm can classify semantic relationships effectively, and the outperform improves by 5 percent.