Research of knowledge mapping construction based on word vector

Vector representations of words learned by the statistical information of word-context matrix encode the semantic meaning of words in semantic space. Knowledge mapping shows the development process and structure of scientific knowledge, using a series of techniques, such as data mining and data visualization. This paper did research on word vectors and proposed a knowledge mapping construction method based on word vectors. Compared with traditional way, the proposed method uses a large amount of contextual information to get the semantic vector representations of keywords. Semantic relation between words can be reflected in the distance between corresponding vectors. Experiments were carried out on the corpus of computer science subject. It is showed that the proposed method was able to mine the field structure of disciplines and to serve as a powerful reference for promoting discipline construction and development. In addition, Comparison with results of traditional method also proved that knowledge mapping constructed by word vectors was indeed more meaningful and unified.