Word’s semantic similarity computation method based on Hownet

Achieved sememe similarity on the basis of hierarchical tree by employing the sememe tree’s depth and density information. Combined the description structure of relation sememe and relation symbol with different weights by analyzing the knowledge description language structure. Proposed an improved semantic similarity computation of Chinese words. This method took the main,secondary and relation features into account,and it could reduce the mistake of regarding the supplement sememe as the basic sememe in secondary features. Experiment shows that the calculating results diffuse to the two poles and become more reasonable,which can distinguish the tiny differences between different words more accurately.