Chinese keyword extraction based on weighted complex network

Abstract

Aiming at the problem of low precision of keyword extraction in traditional complex network method, we propose a keyword extraction method based on an improved weighted complex network, called IWCN algorithm. First, based on the word semantic similarity, we construct a complex network to obtain semantic weight of words. Next, the statistical weight of words is obtained by the introduction of term frequency (TF) and inverse document frequency (IDF). Finally, we combine semantic and statistical weights of words to get keywords. Comparing to traditional complex network approach, the proposed method can avoid the deviations and thus improves extraction accuracy. Simulation results shows that the proposed method achieves higher precision and recall.